{ "data": { "countries": [ { "code": "AR", "name": "Argentina", "usdPerKWh": 0.092 }, { "code": "AU", "name": "Australia", "usdPerKWh": 0.259 }, { "code": "AT", "name": "Austria", "usdPerKWh": 0.356 }, { "code": "BE", "name": "Belgium", "usdPerKWh": 0.409 }, { "code": "BR", "name": "Brazil", "usdPerKWh": 0.166 }, { "code": "CA", "name": "Canada", "usdPerKWh": 0.123 }, { "code": "CN", "name": "China", "usdPerKWh": 0.076 }, { "code": "CZ", "name": "Czechia", "usdPerKWh": 0.35 }, { "code": "DK", "name": "Denmark", "usdPerKWh": 0.352 }, { "code": "FI", "name": "Finland", "usdPerKWh": 0.177 }, { "code": "FR", "name": "France", "usdPerKWh": 0.276 }, { "code": "DE", "name": "Germany", "usdPerKWh": 0.409 }, { "code": "HK", "name": "Hong Kong", "usdPerKWh": 0.184 }, { "code": "IN", "name": "India", "usdPerKWh": 0.076 }, { "code": "ID", "name": "Indonesia", "usdPerKWh": 0.09 }, { "code": "IE", "name": "Ireland", "usdPerKWh": 0.453 }, { "code": "IL", "name": "Israel", "usdPerKWh": 0.188 }, { "code": "IT", "name": "Italy", "usdPerKWh": 0.415 }, { "code": "JP", "name": "Japan", "usdPerKWh": 0.227 }, { "code": "MX", "name": "Mexico", "usdPerKWh": 0.11 }, { "code": "NL", "name": "Netherlands", "usdPerKWh": 0.287 }, { "code": "NZ", "name": "New Zealand", "usdPerKWh": 0.213 }, { "code": "NG", "name": "Nigeria", "usdPerKWh": 0.036 }, { "code": "NO", "name": "Norway", "usdPerKWh": 0.169 }, { "code": "PH", "name": "Philippines", "usdPerKWh": 0.211 }, { "code": "PL", "name": "Poland", "usdPerKWh": 0.236 }, { "code": "PT", "name": "Portugal", "usdPerKWh": 0.24 }, { "code": "RO", "name": "Romania", "usdPerKWh": 0.233 }, { "code": "SG", "name": "Singapore", "usdPerKWh": 0.236 }, { "code": "ZA", "name": "South Africa", "usdPerKWh": 0.215 }, { "code": "KR", "name": "South Korea", "usdPerKWh": 0.126 }, { "code": "ES", "name": "Spain", "usdPerKWh": 0.259 }, { "code": "SE", "name": "Sweden", "usdPerKWh": 0.247 }, { "code": "CH", "name": "Switzerland", "usdPerKWh": 0.367 }, { "code": "TW", "name": "Taiwan", "usdPerKWh": 0.099 }, { "code": "TR", "name": "Türkiye", "usdPerKWh": 0.067 }, { "code": "UA", "name": "Ukraine", "usdPerKWh": 0.085 }, { "code": "AE", "name": "United Arab Emirates", "usdPerKWh": 0.08 }, { "code": "GB", "name": "United Kingdom", "usdPerKWh": 0.405 }, { "code": "US", "name": "United States", "usdPerKWh": 0.19 }, { "code": "VN", "name": "Vietnam", "usdPerKWh": 0.078 } ], "electricityRead": "2026-11-09", "electricitySource": "https://www.globalpetrolprices.com/electricity_prices/", "hardware": [ { "key": "rtx-3060", "name": "NVIDIA RTX 3060 12 GB", "group": "NVIDIA graphics cards", "platform": "nvidia", "memoryGB": 12, "usable": 0.92, "bandwidthGBs": 360, "watts": 170, "match": [ "rtx 3060" ] }, { "key": "rtx-4060-ti-16", "name": "NVIDIA RTX 4060 Ti 16 GB", "group": "NVIDIA graphics cards", "platform": "nvidia", "memoryGB": 16, "usable": 0.92, "bandwidthGBs": 288, "watts": 165, "match": [ "rtx 4060 ti" ] }, { "key": "rtx-3090", "name": "NVIDIA RTX 3090 24 GB", "group": "NVIDIA graphics cards", "platform": "nvidia", "memoryGB": 24, "usable": 0.92, "bandwidthGBs": 936, "watts": 350, "match": [ "rtx 3090" ] }, { "key": "rtx-4090", "name": "NVIDIA RTX 4090 24 GB", "group": "NVIDIA graphics cards", "platform": "nvidia", "memoryGB": 24, "usable": 0.92, "bandwidthGBs": 1008, "watts": 450, "match": [ "rtx 4090" ] }, { "key": "rtx-5070-ti", "name": "NVIDIA RTX 5070 Ti 16 GB", "group": "NVIDIA graphics cards", "platform": "nvidia", "memoryGB": 16, "usable": 0.92, "bandwidthGBs": 896, "watts": 300, "fp4": true, "match": [ "rtx 5070 ti" ] }, { "key": "rtx-5080", "name": "NVIDIA RTX 5080 16 GB", "group": "NVIDIA graphics cards", "platform": "nvidia", "memoryGB": 16, "usable": 0.92, "bandwidthGBs": 960, "watts": 360, "fp4": true, "match": [ "rtx 5080" ] }, { "key": "rtx-5090", "name": "NVIDIA RTX 5090 32 GB", "group": "NVIDIA graphics cards", "platform": "nvidia", "memoryGB": 32, "usable": 0.92, "bandwidthGBs": 1792, "watts": 575, "fp4": true, "match": [ "rtx 5090" ] }, { "key": "rtx-pro-6000", "name": "NVIDIA RTX PRO 6000 Blackwell 96 GB", "group": "NVIDIA graphics cards", "platform": "nvidia", "memoryGB": 96, "usable": 0.92, "bandwidthGBs": 1792, "watts": 600, "fp4": true, "match": [ "rtx pro 6000" ] }, { "key": "m1-m3", "name": "Apple M1, M2 or M3", "group": "Apple silicon", "platform": "apple", "memoryGB": 16, "memoryOptions": [ 8, 16, 24 ], "usable": 0.67, "bandwidthGBs": 100, "watts": 30, "wattsEstimate": true, "match": [ "apple m1", "apple m2", "apple m3" ] }, { "key": "m1-m3-pro", "name": "Apple M1 Pro, M2 Pro or M3 Pro", "group": "Apple silicon", "platform": "apple", "memoryGB": 32, "memoryOptions": [ 16, 18, 32, 36 ], "usable": 0.67, "bandwidthGBs": 150, "watts": 50, "wattsEstimate": true, "match": [ "apple m1 pro", "apple m2 pro", "apple m3 pro" ] }, { "key": "m1-m3-max", "name": "Apple M1 Max, M2 Max or M3 Max", "group": "Apple silicon", "platform": "apple", "memoryGB": 64, "memoryOptions": [ 32, 36, 48, 64, 96, 128 ], "usable": 0.75, "bandwidthGBs": 400, "watts": 80, "wattsEstimate": true, "match": [ "apple m1 max", "apple m2 max", "apple m3 max" ] }, { "key": "m1-m2-ultra", "name": "Apple M1 Ultra or M2 Ultra", "group": "Apple silicon", "platform": "apple", "memoryGB": 128, "memoryOptions": [ 64, 128, 192 ], "usable": 0.75, "bandwidthGBs": 800, "watts": 160, "wattsEstimate": true, "match": [ "apple m1 ultra", "apple m2 ultra" ] }, { "key": "m4", "name": "Apple M4", "group": "Apple silicon", "platform": "apple", "memoryGB": 16, "memoryOptions": [ 16, 24, 32 ], "usable": 0.67, "bandwidthGBs": 120, "watts": 30, "wattsEstimate": true, "match": [ "apple m4" ] }, { "key": "m4-pro", "name": "Apple M4 Pro", "group": "Apple silicon", "platform": "apple", "memoryGB": 24, "memoryOptions": [ 24, 48, 64 ], "usable": 0.67, "bandwidthGBs": 273, "watts": 60, "wattsEstimate": true, "match": [ "apple m4 pro" ] }, { "key": "m4-max", "name": "Apple M4 Max", "group": "Apple silicon", "platform": "apple", "memoryGB": 64, "memoryOptions": [ 36, 48, 64, 128 ], "usable": 0.75, "bandwidthGBs": 546, "watts": 90, "wattsEstimate": true, "match": [ "apple m4 max" ] }, { "key": "m3-ultra", "name": "Apple M3 Ultra", "group": "Apple silicon", "platform": "apple", "memoryGB": 96, "memoryOptions": [ 96, 256, 512 ], "usable": 0.75, "bandwidthGBs": 819, "watts": 180, "wattsEstimate": true, "match": [ "apple m3 ultra" ] }, { "key": "m5", "name": "Apple M5", "group": "Apple silicon", "platform": "apple", "memoryGB": 16, "memoryOptions": [ 16, 24, 32 ], "usable": 0.67, "bandwidthGBs": 153, "watts": 30, "wattsEstimate": true, "match": [ "apple m5" ] }, { "key": "m5-pro", "name": "Apple M5 Pro", "group": "Apple silicon", "platform": "apple", "memoryGB": 48, "memoryOptions": [ 24, 48, 64 ], "usable": 0.67, "bandwidthGBs": 307, "watts": 65, "wattsEstimate": true, "match": [ "apple m5 pro" ] }, { "key": "m5-max", "name": "Apple M5 Max", "group": "Apple silicon", "platform": "apple", "memoryGB": 64, "memoryOptions": [ 36, 48, 64, 128 ], "usable": 0.75, "bandwidthGBs": 614, "watts": 100, "wattsEstimate": true, "match": [ "apple m5 max" ] }, { "key": "strix-halo", "name": "AMD Ryzen AI Max+ 395 (Strix Halo)", "group": "Unified-memory PCs", "platform": "amd-unified", "memoryGB": 128, "memoryOptions": [ 64, 96, 128 ], "usable": 0.75, "bandwidthGBs": 256, "watts": 140, "wattsEstimate": true, "match": [ "radeon 8060s", "radeon(tm) 8060s" ] }, { "key": "dgx-spark", "name": "NVIDIA DGX Spark", "group": "Unified-memory PCs", "platform": "nvidia-unified", "memoryGB": 128, "usable": 0.85, "bandwidthGBs": 273, "watts": 200, "wattsEstimate": true, "fp4": true, "match": [ "gb10" ] }, { "key": "cpu", "name": "Any computer, on the CPU", "group": "No graphics card", "platform": "cpu", "memoryGB": 16, "memoryOptions": [ 8, 16, 32, 64, 128 ], "usable": 0.6, "bandwidthGBs": 80, "watts": 100, "wattsEstimate": true } ], "method": { "assumption": "Nothing the browser reports is sent to us. A browser can't tell how much memory a machine has, so memory is a guess to correct.", "cost": "Power under load in watts times hours a day times 30, in kWh, times the electricity price. A million tokens takes a million divided by tokens a second, in hours, at the same power.", "fits": "A build fits when its memoryGB (the file plus a fifth for the runtime and a working context) is at most the machine's memory times its usable share.", "pick": "The best-scored listing for the task with a build that fits, and of its fitting builds the 8-bit one, then 6, 5 and 4-bit, and 16 or 32-bit only when nothing quantised fits.", "platforms": "NVIDIA cards run every format but MLX and Core ML, and NVFP4 builds only on Blackwell. Apple silicon runs GGUF, Ollama, MLX, Core ML, ONNX and unquantised safetensors. AMD unified-memory PCs and plain CPUs run GGUF, Ollama, ONNX and unquantised safetensors, and AMD also the builds that name ROCm.", "speed": "Tokens a second for one stream, an estimate: memory bandwidth times an efficiency (0.6 for a graphics card, 0.65 for Apple silicon and unified memory, 0.5 for a CPU) divided by the gigabytes read for each token, the build's file size times the active share of its parameters." }, "models": [ { "slug": "qwen3-embedding-8b", "name": "Qwen3-Embedding-8B", "grade": "BB", "score": 75.4, "tasks": [ "embed" ], "paramsB": 7.57, "params": "7.57B", "context": 32768, "goodFor": "Multilingual text and code embeddings for retrieval, clustering and classification, with task instructions and vectors that can be shortened.", "licence": "Apache-2.0", "run": "llama-server -hf Qwen/Qwen3-Embedding-8B-GGUF:Q4_K_M --embedding --pooling last -ub 8192", "runBuild": "Qwen3-Embedding-8B-Q4_K_M", "url": "/tools/qwen3-embedding-8b", "builds": [ { "name": "Qwen3-Embedding-8B-Q6_K", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-8B-GGUF/blob/main/Qwen3-Embedding-8B-Q6_K.gguf", "precision": "q6_k", "format": "gguf", "sizeGB": 6.2, "memoryGB": 8 }, { "name": "Qwen3-Embedding-8B-Q5_K_M", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-8B-GGUF/blob/main/Qwen3-Embedding-8B-Q5_K_M.gguf", "precision": "q5_k_m", "format": "gguf", "sizeGB": 5.4, "memoryGB": 8 }, { "name": "Qwen3-Embedding-8B-Q5_0", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-8B-GGUF/blob/main/Qwen3-Embedding-8B-Q5_0.gguf", "precision": "q5_0", "format": "gguf", "sizeGB": 5.3, "memoryGB": 8 }, { "name": "Qwen3-Embedding-8B-Q4_K_M", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-8B-GGUF/blob/main/Qwen3-Embedding-8B-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 4.7, "memoryGB": 8 }, { "name": "ollama qwen3-embedding:8b", "url": "https://ollama.com/library/qwen3-embedding:8b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 4.7, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "Qwen3-Embedding-8B-Q8_0", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-8B-GGUF/blob/main/Qwen3-Embedding-8B-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 8, "memoryGB": 12 }, { "name": "ollama qwen3-embedding:8b-q8_0", "url": "https://ollama.com/library/qwen3-embedding:8b-q8_0", "precision": "q8_0", "format": "ollama", "sizeGB": 8, "memoryGB": 12, "runsOn": [ "Ollama" ] }, { "name": "Qwen3-Embedding-8B", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-8B", "precision": "bf16", "format": "safetensors", "sizeGB": 15.2, "memoryGB": 24 }, { "name": "Qwen3-Embedding-8B-f16", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-8B-GGUF/blob/main/Qwen3-Embedding-8B-f16.gguf", "precision": "f16", "format": "gguf", "sizeGB": 15.1, "memoryGB": 24 }, { "name": "ollama qwen3-embedding:8b-fp16", "url": "https://ollama.com/library/qwen3-embedding:8b-fp16", "precision": "fp16", "format": "ollama", "sizeGB": 15, "memoryGB": 24, "runsOn": [ "Ollama" ] } ] }, { "slug": "gemma-4-31b-it", "name": "Gemma 4 31B IT", "grade": "BB", "score": 74.1, "tasks": [ "agents", "general", "vision" ], "paramsB": 30.7, "params": "30.7B", "context": 262144, "goodFor": "Dense general assistant and agent model with tool calling and optional thinking, taking text, images and video frames.", "licence": "Apache-2.0", "run": "llama-server -hf google/gemma-4-31B-it-qat-q4_0-gguf --hf-file gemma-4-31B_q4_0-it.gguf --jinja -c 32768", "runBuild": "gemma-4-31B_q4_0-it", "url": "/tools/gemma-4-31b-it", "builds": [ { "name": "gemma-4-31B_q4_0-it", "url": "https://huggingface.co/google/gemma-4-31B-it-qat-q4_0-gguf/blob/main/gemma-4-31B_q4_0-it.gguf", "precision": "q4_0", "format": "gguf", "sizeGB": 17.7, "memoryGB": 24, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)", "LM Studio" ] }, { "name": "ollama gemma4:31b", "url": "https://ollama.com/library/gemma4:31b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 20, "memoryGB": 24, "runsOn": [ "Ollama" ] }, { "name": "gemma-4-31B-it-qat-w4a16-ct", "url": "https://huggingface.co/google/gemma-4-31B-it-qat-w4a16-ct", "precision": "w4a16", "format": "safetensors", "sizeGB": 23.3, "memoryGB": 32 }, { "name": "gemma-4-31B-it", "url": "https://huggingface.co/google/gemma-4-31B-it", "precision": "bf16", "format": "safetensors", "sizeGB": 62.5, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "qwen3-embedding-4b", "name": "Qwen3-Embedding-4B", "grade": "BB", "score": 73.6, "tasks": [ "embed" ], "paramsB": 4.02, "params": "4.02B", "context": 32768, "goodFor": "Multilingual text and code embeddings for retrieval, clustering and classification on small hardware, with task instructions and vectors that can be shortened.", "licence": "Apache-2.0", "run": "llama-server -hf Qwen/Qwen3-Embedding-4B-GGUF:Q4_K_M --embedding --pooling last -ub 8192", "runBuild": "Qwen3-Embedding-4B-Q4_K_M", "url": "/tools/qwen3-embedding-4b", "builds": [ { "name": "Qwen3-Embedding-4B-Q8_0", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-4B-GGUF/blob/main/Qwen3-Embedding-4B-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 4.3, "memoryGB": 8 }, { "name": "Qwen3-Embedding-4B-Q6_K", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-4B-GGUF/blob/main/Qwen3-Embedding-4B-Q6_K.gguf", "precision": "q6_k", "format": "gguf", "sizeGB": 3.3, "memoryGB": 8 }, { "name": "Qwen3-Embedding-4B-Q5_K_M", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-4B-GGUF/blob/main/Qwen3-Embedding-4B-Q5_K_M.gguf", "precision": "q5_k_m", "format": "gguf", "sizeGB": 2.9, "memoryGB": 8 }, { "name": "Qwen3-Embedding-4B-Q5_0", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-4B-GGUF/blob/main/Qwen3-Embedding-4B-Q5_0.gguf", "precision": "q5_0", "format": "gguf", "sizeGB": 2.8, "memoryGB": 8 }, { "name": "Qwen3-Embedding-4B-Q4_K_M", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-4B-GGUF/blob/main/Qwen3-Embedding-4B-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 2.5, "memoryGB": 8 }, { "name": "ollama qwen3-embedding:4b", "url": "https://ollama.com/library/qwen3-embedding:4b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 2.5, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "ollama qwen3-embedding:4b-q8_0", "url": "https://ollama.com/library/qwen3-embedding:4b-q8_0", "precision": "q8_0", "format": "ollama", "sizeGB": 4.3, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "Qwen3-Embedding-4B", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-4B", "precision": "bf16", "format": "safetensors", "sizeGB": 8.1, "memoryGB": 12 }, { "name": "Qwen3-Embedding-4B-f16", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-4B-GGUF/blob/main/Qwen3-Embedding-4B-f16.gguf", "precision": "f16", "format": "gguf", "sizeGB": 8, "memoryGB": 12 }, { "name": "ollama qwen3-embedding:4b-fp16", "url": "https://ollama.com/library/qwen3-embedding:4b-fp16", "precision": "fp16", "format": "ollama", "sizeGB": 8, "memoryGB": 12, "runsOn": [ "Ollama" ] } ] }, { "slug": "gemma-4-12b-it", "name": "Gemma 4 12B IT", "grade": "BB", "score": 72.3, "tasks": [ "agents", "general", "vision" ], "paramsB": 11.95, "params": "11.95B", "context": 262144, "goodFor": "Mid-size dense assistant and agent model for a laptop GPU, with tool calling and optional thinking, taking text, images, video frames and audio without separate encoders, and a 256K context.", "licence": "Apache-2.0", "run": "llama-server -hf google/gemma-4-12B-it-qat-q4_0-gguf --hf-file gemma-4-12b-it-qat-q4_0.gguf --jinja -c 131072", "runBuild": "gemma-4-12b-it-qat-q4_0", "url": "/tools/gemma-4-12b-it", "builds": [ { "name": "gemma-4-12b-it-qat-q4_0", "url": "https://huggingface.co/google/gemma-4-12B-it-qat-q4_0-gguf/blob/main/gemma-4-12b-it-qat-q4_0.gguf", "precision": "q4_0", "format": "gguf", "sizeGB": 7.15, "memoryGB": 12, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)", "LM Studio" ] }, { "name": "ollama gemma4:12b", "url": "https://ollama.com/library/gemma4:12b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 8, "memoryGB": 12, "runsOn": [ "Ollama" ] }, { "name": "lmstudio google/gemma-4-12b", "url": "https://lmstudio.ai/models/google/gemma-4-12b", "precision": "q4_k_m", "format": "gguf", "sizeGB": 7.4, "memoryGB": 12, "runsOn": [ "LM Studio" ] }, { "name": "gemma-4-12B-it-qat-w4a16-ct", "url": "https://huggingface.co/google/gemma-4-12B-it-qat-w4a16-ct", "precision": "w4a16", "format": "safetensors", "sizeGB": 10.3, "memoryGB": 16 }, { "name": "gemma-4-12B-it", "url": "https://huggingface.co/google/gemma-4-12B-it", "precision": "bf16", "format": "safetensors", "sizeGB": 23.95, "memoryGB": 32, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "gemma-4-26b-a4b-it", "name": "Gemma 4 26B A4B IT", "grade": "BB", "score": 72.2, "tasks": [ "agents", "general", "vision" ], "paramsB": 25.2, "activeB": 3.8, "params": "25.2B", "context": 262144, "goodFor": "General assistant and agent model with tool calling and optional thinking, taking text, images and video frames, fast for its size as a mixture of experts.", "licence": "Apache-2.0", "run": "llama-server -hf google/gemma-4-26B-A4B-it-qat-q4_0-gguf --hf-file gemma-4-26B_q4_0-it.gguf --jinja -c 65536", "runBuild": "gemma-4-26B_q4_0-it", "url": "/tools/gemma-4-26b-a4b-it", "builds": [ { "name": "gemma-4-26B_q4_0-it", "url": "https://huggingface.co/google/gemma-4-26B-A4B-it-qat-q4_0-gguf/blob/main/gemma-4-26B_q4_0-it.gguf", "precision": "q4_0", "format": "gguf", "sizeGB": 14.4, "memoryGB": 24, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)", "LM Studio" ] }, { "name": "ollama gemma4:26b", "url": "https://ollama.com/library/gemma4:26b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 19, "memoryGB": 24, "runsOn": [ "Ollama" ] }, { "name": "gemma-4-26B-A4B-it", "url": "https://huggingface.co/google/gemma-4-26B-A4B-it", "precision": "bf16", "format": "safetensors", "sizeGB": 51.6, "memoryGB": 64, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "gpt-oss-120b", "name": "gpt-oss-120b", "grade": "BB", "score": 70.2, "tasks": [ "agents", "general" ], "paramsB": 116.83, "activeB": 5.13, "params": "116.83B", "context": 131072, "goodFor": "Larger text reasoning model for agents and coding, with function calling, built-in browser and Python tools and three reasoning levels, in the harmony format.", "licence": "Apache-2.0", "run": "ollama run gpt-oss:120b", "runBuild": "ollama gpt-oss:120b", "url": "/tools/gpt-oss-120b", "builds": [ { "name": "gpt-oss-120b", "url": "https://huggingface.co/openai/gpt-oss-120b", "precision": "mxfp4", "format": "safetensors", "sizeGB": 65.2, "memoryGB": 96 }, { "name": "ollama gpt-oss:120b", "url": "https://ollama.com/library/gpt-oss:120b", "precision": "mxfp4", "format": "ollama", "sizeGB": 65, "memoryGB": 96, "runsOn": [ "Ollama" ] } ] }, { "slug": "muse-glimmer-30b", "name": "Muse Glimmer 30B", "grade": "BB", "score": 70, "tasks": [ "agents", "general", "vision" ], "paramsB": 29.6, "params": "29.6B", "context": 131072, "goodFor": "Dense multimodal model for local agent work, text and images in, with tool calls and always-on adjustable reasoning.", "licence": "Apache-2.0", "run": "llama serve -hf meta-models/Muse-Glimmer-30B-GGUF --hf-file Muse-Glimmer-30B-KQuant-17GB-Q4_K_M.gguf -ngl 99 -c 131072 --host 127.0.0.1 --port 8080 --chat-template-kwargs '{\"reasoning_strength\":\"high\"}'", "runBuild": "Muse-Glimmer-30B-KQuant-17GB-Q4_K_M", "url": "/tools/muse-glimmer-30b", "builds": [ { "name": "Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL", "url": "https://huggingface.co/meta-models/Muse-Glimmer-30B-GGUF/blob/main/Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf", "precision": "q4_k_xl", "format": "gguf", "sizeGB": 19.7, "memoryGB": 24, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)", "LM Studio" ] }, { "name": "Muse-Glimmer-30B-KQuant-17GB-Q4_K_M", "url": "https://huggingface.co/meta-models/Muse-Glimmer-30B-GGUF/blob/main/Muse-Glimmer-30B-KQuant-17GB-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 16.8, "memoryGB": 24, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)", "LM Studio" ] }, { "name": "Muse-Glimmer-30B-ExecuTorch-PTE (smallest: k-quant-17G text-solo, Metal)", "url": "https://huggingface.co/meta-models/Muse-Glimmer-30B-ExecuTorch-PTE", "precision": "q4_k_m", "format": "executorch", "sizeGB": 17.9, "memoryGB": 24 }, { "name": "ollama muse-glimmer:latest", "url": "https://ollama.com/library/muse-glimmer:latest", "precision": "q4_k_m", "format": "ollama", "sizeGB": 18, "memoryGB": 24, "runsOn": [ "Ollama" ] }, { "name": "Muse-Glimmer-30B", "url": "https://huggingface.co/meta-models/Muse-Glimmer-30B", "precision": "bf16", "format": "safetensors", "sizeGB": 59.6, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "gpt-oss-20b", "name": "gpt-oss-20b", "grade": "B", "score": 69.9, "tasks": [ "agents", "general" ], "paramsB": 20.91, "activeB": 3.61, "params": "20.91B", "context": 131072, "goodFor": "Text reasoning model for agents and coding, with function calling, built-in browser and Python tools and low, medium or high reasoning, in the harmony format.", "licence": "Apache-2.0", "run": "ollama run gpt-oss:20b", "runBuild": "ollama gpt-oss:20b", "url": "/tools/gpt-oss-20b", "builds": [ { "name": "gpt-oss-20b", "url": "https://huggingface.co/openai/gpt-oss-20b", "precision": "mxfp4", "format": "safetensors", "sizeGB": 13.8, "memoryGB": 24 }, { "name": "ollama gpt-oss:20b", "url": "https://ollama.com/library/gpt-oss:20b", "precision": "mxfp4", "format": "ollama", "sizeGB": 14, "memoryGB": 24, "runsOn": [ "Ollama" ] } ] }, { "slug": "nemotron-3-ultra-550b-a55b", "name": "Nemotron 3 Ultra 550B A55B", "grade": "B", "score": 69.3, "tasks": [ "agents", "general" ], "paramsB": 550, "activeB": 55, "params": "550B", "context": 1048576, "goodFor": "Frontier-scale hybrid Mamba and latent mixture-of-experts reasoning model for long-running agents, text only, in ten languages, with tool calling, a medium-effort mode and up to 1M tokens of context.", "licence": "OpenMDW-1.1", "run": "vllm serve nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4 --served-model-name nvidia/nemotron-3-ultra --tensor-parallel-size 4 --trust-remote-code --kv-cache-dtype fp8 --max-model-len 262144 --reasoning-parser nemotron_v3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4", "url": "/tools/nemotron-3-ultra-550b-a55b", "builds": [ { "name": "NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 352.38, "memoryGB": 512 }, { "name": "NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16", "precision": "bf16", "format": "safetensors", "sizeGB": 1121.08, "memoryGB": 2048 } ] }, { "slug": "nemotron-3-5-parse", "name": "Nemotron 3.5 Parse", "grade": "B", "score": 69, "tasks": [ "docs" ], "paramsB": 0.63, "params": "0.63B", "context": 16384, "goodFor": "Turning scanned or rendered document pages into Markdown, HTML tables and JSON blocks with layout classes, boxes and reading order, fast on one GPU, for retrieval, extraction and data-curation pipelines.", "licence": "OpenMDW-1.1", "run": "hf download nvidia/Nemotron-3.5-Parse --local-dir nemotron-parse \u0026\u0026 cd nemotron-parse \u0026\u0026 pip install \".[vllm,client]\" \"vllm\u003e=0.25.0,\u003c0.33\" \u0026\u0026 vllm serve . --served-model-name nemotron-parse --dtype bfloat16 --trust-remote-code --max-model-len 16384 --no-enable-prefix-caching --limit-mm-per-prompt '{\"image\":1}' --chat-template chat_template.jinja", "runBuild": "Nemotron-3.5-Parse (safetensors)", "url": "/tools/nemotron-3-5-parse", "builds": [ { "name": "Nemotron-3.5-Parse (safetensors)", "url": "https://huggingface.co/nvidia/Nemotron-3.5-Parse/blob/main/model.safetensors", "precision": "fp32", "format": "safetensors", "sizeGB": 2.52, "memoryGB": 8, "runsOn": [ "Transformers (remote code)", "vLLM (NVIDIA's plugin)" ] } ] }, { "slug": "qwen3-8-27b", "name": "Qwen3.8-27B", "grade": "B", "score": 68.6, "tasks": [ "agents", "general", "vision" ], "paramsB": 27, "params": "27B", "context": 262144, "goodFor": "Dense vision-language model for coding and agents, text, images and video in, tool calling and adjustable reasoning effort.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3.8-27B-FP8 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "Qwen3.8-27B-FP8", "url": "/tools/qwen3-8-27b", "builds": [ { "name": "ollama qwen3.8:latest", "url": "https://ollama.com/library/qwen3.8:latest", "precision": "q4_k_m", "format": "ollama", "sizeGB": 18, "memoryGB": 24, "runsOn": [ "Ollama" ] }, { "name": "Qwen3.8-27B-FP8", "url": "https://huggingface.co/Qwen/Qwen3.8-27B-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 30.9, "memoryGB": 48 }, { "name": "Qwen3.8-27B", "url": "https://huggingface.co/Qwen/Qwen3.8-27B", "precision": "bf16", "format": "safetensors", "sizeGB": 55.6, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "nemotron-3-nano-30b-a3b", "name": "Nemotron 3 Nano 30B A3B", "grade": "B", "score": 67.6, "tasks": [ "agents", "general" ], "paramsB": 30, "activeB": 3.5, "params": "30B", "context": 1048576, "goodFor": "Hybrid Mamba-Transformer mixture of experts for reasoning and agent work, text only, with tool calling and a reasoning budget.", "licence": "NVIDIA Nemotron Open Model License (custom, Apache-style, commercial use allowed, not an OSI licence)", "run": "vllm serve nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-NVFP4 --max-model-len 262144 --trust-remote-code --enable-auto-tool-choice --tool-call-parser qwen3_coder --reasoning-parser-plugin nano_v3_reasoning_parser.py --reasoning-parser nano_v3", "runBuild": "NVIDIA-Nemotron-3-Nano-30B-A3B-NVFP4", "url": "/tools/nemotron-3-nano-30b-a3b", "builds": [ { "name": "NVIDIA-Nemotron-3-Nano-30B-A3B-NVFP4", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 19.3, "memoryGB": 24 }, { "name": "ollama nemotron-3-nano:30b", "url": "https://ollama.com/library/nemotron-3-nano:30b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 24, "memoryGB": 32, "runsOn": [ "Ollama" ] }, { "name": "NVIDIA-Nemotron-3-Nano-30B-A3B-FP8", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 32.7, "memoryGB": 48 }, { "name": "NVIDIA-Nemotron-3-Nano-30B-A3B-BF16", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16", "precision": "bf16", "format": "safetensors", "sizeGB": 63.2, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "inkling-small", "name": "Inkling-Small", "grade": "B", "score": 67, "tasks": [ "agents", "general", "vision" ], "paramsB": 276, "activeB": 12, "params": "276B", "context": 1048576, "goodFor": "Mixture-of-experts generalist for coding, agents and tool use with images and audio in, and thinking effort set per request.", "licence": "Apache-2.0", "run": "sglang serve --trust-remote-code --model-path thinkingmachines/Inkling-Small-NVFP4 --tp 8 --enable-multimodal --reasoning-parser inkling --tool-call-parser inkling", "runBuild": "Inkling-Small-NVFP4", "url": "/tools/inkling-small", "builds": [ { "name": "Inkling-Small-NVFP4", "url": "https://huggingface.co/thinkingmachines/Inkling-Small-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 170.8, "memoryGB": 256 }, { "name": "Inkling-Small", "url": "https://huggingface.co/thinkingmachines/Inkling-Small", "precision": "bf16", "format": "safetensors", "sizeGB": 531.9, "memoryGB": 1024 } ] }, { "slug": "parakeet-tdt-0-6b-v3", "name": "Parakeet TDT 0.6B v3", "grade": "B", "score": 66.9, "tasks": [ "stt" ], "paramsB": 0.6, "params": "0.6B", "goodFor": "Fast multilingual transcription in 25 European languages with automatic language detection, punctuation and word-level timestamps.", "licence": "CC-BY-4.0", "run": "nemo-speech transcribe speech.wav --model models/parakeet-tdt-0.6b-v3.q8_0.gguf", "runBuild": "parakeet-tdt-0.6b-v3.q8_0.gguf", "url": "/tools/parakeet-tdt-0-6b-v3", "builds": [ { "name": "parakeet-tdt-0.6b-v3 (safetensors)", "url": "https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3/blob/main/model.safetensors", "precision": "fp32", "format": "safetensors", "sizeGB": 2.5, "memoryGB": 8, "runsOn": [ "Transformers" ] }, { "name": "parakeet-tdt-0.6b-v3.nemo", "url": "https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3/blob/main/parakeet-tdt-0.6b-v3.nemo", "precision": "fp32", "format": "nemo", "sizeGB": 2.5, "memoryGB": 8, "runsOn": [ "NeMo" ] }, { "name": "parakeet-tdt-0.6b-v3.q8_0.gguf", "url": "https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3/blob/main/parakeet-tdt-0.6b-v3.q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 0.71, "memoryGB": 8, "runsOn": [ "NeMo-Speech.cpp" ] } ] }, { "slug": "nemotron-3-super-120b-a12b", "name": "Nemotron 3 Super 120B A12B", "grade": "B", "score": 66.7, "tasks": [ "agents", "general" ], "paramsB": 120, "activeB": 12, "params": "120B", "context": 1048576, "goodFor": "Hybrid Mamba and mixture-of-experts reasoning model for agents, text only, with tool calling, a low-effort mode and a reasoning budget.", "licence": "NVIDIA Nemotron Open Model License", "run": "vllm serve nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4 --served-model-name nvidia/nemotron-3-super --max-model-len 262144 --kv-cache-dtype fp8 --mamba-ssm-cache-dtype float32 --trust-remote-code --reasoning-parser nemotron_v3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4", "url": "/tools/nemotron-3-super-120b-a12b", "builds": [ { "name": "NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 80.3, "memoryGB": 128 }, { "name": "ollama nemotron-3-super:latest", "url": "https://ollama.com/library/nemotron-3-super:latest", "precision": "q4_k_m", "format": "ollama", "sizeGB": 87, "memoryGB": 128, "runsOn": [ "Ollama" ] }, { "name": "NVIDIA-Nemotron-3-Super-120B-A12B-FP8", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 128.4, "memoryGB": 192 }, { "name": "NVIDIA-Nemotron-3-Super-120B-A12B-BF16", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16", "precision": "bf16", "format": "safetensors", "sizeGB": 247.2, "memoryGB": 512, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "qwen3-6-27b", "name": "Qwen3.6-27B", "grade": "B", "score": 66.7, "tasks": [ "agents", "general", "coding", "vision" ], "paramsB": 27, "params": "27B", "context": 262144, "goodFor": "Dense vision-language model for agentic coding and tool calling, text, images and video in, thinking on by default with reasoning kept across turns if asked.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3.6-27B-FP8 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "Qwen3.6-27B-FP8", "url": "/tools/qwen3-6-27b", "builds": [ { "name": "ollama qwen3.6:27b", "url": "https://ollama.com/library/qwen3.6:27b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 18, "memoryGB": 24, "runsOn": [ "Ollama" ] }, { "name": "LM Studio qwen/qwen3.6-27b (Q4_K_M GGUF)", "url": "https://huggingface.co/lmstudio-community/Qwen3.6-27B-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 17.5, "memoryGB": 24, "runsOn": [ "LM Studio", "llama.cpp" ] }, { "name": "Qwen3.6-27B-FP8", "url": "https://huggingface.co/Qwen/Qwen3.6-27B-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 30.9, "memoryGB": 48 }, { "name": "Qwen3.6-27B", "url": "https://huggingface.co/Qwen/Qwen3.6-27B", "precision": "bf16", "format": "safetensors", "sizeGB": 55.6, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "embeddinggemma-2", "name": "EmbeddingGemma 2", "grade": "B", "score": 66.4, "tasks": [ "embed" ], "paramsB": 0.74, "params": "740M", "context": 8192, "goodFor": "Small multimodal embeddings for search, retrieval, classification and clustering across text, code, images, video and audio in one vector space, on laptops, phones or in a browser.", "licence": "Apache 2.0", "run": "python -c \"from sentence_transformers import SentenceTransformer; SentenceTransformer('google/embeddinggemma-2')\"", "runBuild": "embeddinggemma-2", "url": "/tools/embeddinggemma-2", "builds": [ { "name": "embeddinggemma-2", "url": "https://huggingface.co/google/embeddinggemma-2", "precision": "bf16", "format": "safetensors", "sizeGB": 1.53, "memoryGB": 8, "runsOn": [ "Transformers", "vLLM", "SGLang" ] }, { "name": "ggml-org embeddinggemma-2-Q8_0", "url": "https://huggingface.co/ggml-org/embeddinggemma-2-GGUF", "precision": "q8_0", "format": "gguf", "sizeGB": 0.86, "memoryGB": 8, "runsOn": [ "llama.cpp" ] }, { "name": "onnx-community embeddinggemma-2-ONNX q4", "url": "https://huggingface.co/onnx-community/embeddinggemma-2-ONNX", "precision": "q4", "format": "onnx", "sizeGB": 0.5, "memoryGB": 8, "runsOn": [ "Transformers.js" ] }, { "name": "litert-community embeddinggemma-2-740m-litert-lm", "url": "https://huggingface.co/litert-community/embeddinggemma-2-740m-litert-lm", "precision": "int4 (QAT, int8 vision, mixed audio)", "format": "litertlm", "sizeGB": 0.48, "memoryGB": 8, "runsOn": [ "LiteRT-LM" ] } ] }, { "slug": "glm-5-3-flash", "name": "GLM-5.3-Flash", "grade": "B", "score": 65.1, "tasks": [ "agents", "general", "vision" ], "paramsB": 320, "activeB": 18, "params": "320B", "context": 1048576, "goodFor": "Multimodal agent and coding model: text, images and video in, text out, with tool calling and always-on reasoning.", "licence": "MIT", "run": "vllm serve zai-org/GLM-5.3-Flash --tensor-parallel-size 4 --enable-auto-tool-choice --tool-call-parser glm47 --reasoning-parser glm47", "runBuild": "GLM-5.3-Flash", "url": "/tools/glm-5-3-flash", "builds": [ { "name": "GLM-5.3-Flash-FP4-Local", "url": "https://huggingface.co/nvidia/GLM-5.3-Flash-FP4-Local", "precision": "nvfp4", "format": "safetensors", "sizeGB": 177.5, "memoryGB": 256 }, { "name": "GLM-5.3-Flash-NVFP4", "url": "https://huggingface.co/nvidia/GLM-5.3-Flash-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 204.4, "memoryGB": 256 }, { "name": "GLM-5.3-Flash", "url": "https://huggingface.co/zai-org/GLM-5.3-Flash", "precision": "fp8", "format": "safetensors", "sizeGB": 328.3, "memoryGB": 512 }, { "name": "GLM-5.3-Flash-BF16", "url": "https://huggingface.co/zai-org/GLM-5.3-Flash-BF16", "precision": "bf16", "format": "safetensors", "sizeGB": 642.7, "memoryGB": 1024, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "inkling", "name": "Inkling", "grade": "B", "score": 64.7, "tasks": [ "agents", "general", "vision" ], "paramsB": 975, "activeB": 41, "params": "975B", "context": 1048576, "goodFor": "Large mixture-of-experts generalist for coding, agents, factual recall and audio, with images in and thinking effort set per request.", "licence": "Apache-2.0", "run": "sglang serve --trust-remote-code --model-path thinkingmachines/Inkling-NVFP4 --tp 8 --quantization modelopt_fp4 --enable-multimodal --reasoning-parser inkling --tool-call-parser inkling", "runBuild": "Inkling-NVFP4", "url": "/tools/inkling", "builds": [ { "name": "Inkling-NVFP4", "url": "https://huggingface.co/thinkingmachines/Inkling-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 592, "memoryGB": 1024 }, { "name": "Inkling", "url": "https://huggingface.co/thinkingmachines/Inkling", "precision": "bf16", "format": "safetensors", "sizeGB": 1904.8, "memoryGB": 4096 } ] }, { "slug": "nemotron-3-5-lightning-30b-a3b", "name": "Nemotron 3.5 Lightning 30B A3B", "grade": "B", "score": 64.7, "tasks": [ "agents", "general" ], "paramsB": 30, "activeB": 3, "params": "30B", "context": 1048576, "goodFor": "Small hybrid Mamba and mixture-of-experts reasoning model for always-on agents, text only, with tool calling, a thinking switch and up to 1M tokens of context.", "licence": "OpenMDW-1.1", "run": "llama-server -hf ggml-org/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:Q4_0 --jinja --temp 1.0 --top-p 0.95 -c 131072", "runBuild": "ggml-org NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF (Q4_0)", "url": "/tools/nemotron-3-5-lightning-30b-a3b", "builds": [ { "name": "ggml-org NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF (Q4_0)", "url": "https://huggingface.co/ggml-org/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF", "precision": "q4_0", "format": "gguf", "sizeGB": 18.9, "memoryGB": 24, "runsOn": [ "llama.cpp" ] }, { "name": "ollama nemotron-3.5-lightning:30b-a3b-nvfp4", "url": "https://ollama.com/library/nemotron-3.5-lightning:30b-a3b-nvfp4", "precision": "nvfp4", "format": "ollama", "sizeGB": 19, "memoryGB": 24, "runsOn": [ "Ollama (MLX)" ] }, { "name": "NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 21.58, "memoryGB": 32 }, { "name": "ollama nemotron-3.5-lightning:30b-a3b-q4_K_M", "url": "https://ollama.com/library/nemotron-3.5-lightning:30b-a3b-q4_K_M", "precision": "q4_k_m", "format": "ollama", "sizeGB": 25, "memoryGB": 32, "runsOn": [ "Ollama" ] }, { "name": "lmstudio-community NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF (Q4_K_M)", "url": "https://huggingface.co/lmstudio-community/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 24.52, "memoryGB": 32, "runsOn": [ "LM Studio" ] }, { "name": "NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16", "precision": "bf16", "format": "safetensors", "sizeGB": 65.85, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "glm-5-3", "name": "GLM-5.3", "grade": "B", "score": 64.6, "tasks": [ "agents", "general", "coding" ], "paramsB": 744, "activeB": 40, "params": "744B", "context": 1048576, "goodFor": "Large model for coding and long-horizon agent work with tool calling and always-on reasoning, text only.", "licence": "GLM-5.3 License (MIT-style, with a security review for the largest model-as-a-service operators)", "run": "vllm serve zai-org/GLM-5.3 --kv-cache-dtype fp8 --tensor-parallel-size 8 --speculative-config.method mtp --speculative-config.num_speculative_tokens 5 --tool-call-parser glm47 --reasoning-parser glm47 --enable-auto-tool-choice --served-model-name glm-5.3", "runBuild": "GLM-5.3", "url": "/tools/glm-5-3", "builds": [ { "name": "GLM-5.3", "url": "https://huggingface.co/zai-org/GLM-5.3", "precision": "fp8", "format": "safetensors", "sizeGB": 755.6, "memoryGB": 1024 }, { "name": "GLM-5.3-BF16", "url": "https://huggingface.co/zai-org/GLM-5.3-BF16", "precision": "bf16", "format": "safetensors", "sizeGB": 1506.7, "memoryGB": 2048, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "qwen3-asr-1-7b", "name": "Qwen3-ASR-1.7B", "grade": "B", "score": 64.4, "tasks": [ "stt" ], "paramsB": 2.04, "params": "2.04B", "context": 65536, "goodFor": "Multilingual and Chinese-dialect transcription with language identification, context hints and streaming, timestamps through the companion forced aligner.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3-ASR-1.7B", "runBuild": "Qwen3-ASR-1.7B", "url": "/tools/qwen3-asr-1-7b", "builds": [ { "name": "Qwen3-ASR-1.7B", "url": "https://huggingface.co/Qwen/Qwen3-ASR-1.7B", "precision": "bf16", "format": "safetensors", "sizeGB": 4.7, "memoryGB": 8, "runsOn": [ "vLLM", "SGLang" ] }, { "name": "Qwen3-ASR-1.7B-hf", "url": "https://huggingface.co/Qwen/Qwen3-ASR-1.7B-hf", "precision": "bf16", "format": "safetensors", "sizeGB": 4.09, "memoryGB": 8, "runsOn": [ "Transformers" ] }, { "name": "ggml-org Qwen3-ASR-1.7B-GGUF Q8_0", "url": "https://huggingface.co/ggml-org/Qwen3-ASR-1.7B-GGUF", "precision": "q8_0", "format": "gguf", "sizeGB": 2.52, "memoryGB": 8, "runsOn": [ "llama.cpp" ] } ] }, { "slug": "qwen3-5-397b-a17b", "name": "Qwen3.5-397B-A17B", "grade": "B", "score": 64.1, "tasks": [ "agents", "general", "vision" ], "params": "397B total, 17B active", "context": 262144, "goodFor": "Alibaba's largest open Qwen3.5, a mixture-of-experts vision-language model for coding, agents and tool calling on a multi-GPU server, thinking on by default.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3.5-397B-A17B-FP8 --port 8000 --tensor-parallel-size 8 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "Qwen3.5-397B-A17B-FP8", "url": "/tools/qwen3-5-397b-a17b", "builds": [ { "name": "Qwen3.5-397B-A17B-FP8", "url": "https://huggingface.co/Qwen/Qwen3.5-397B-A17B-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 406.2, "memoryGB": 512 }, { "name": "Qwen3.5-397B-A17B-GPTQ-Int4", "url": "https://huggingface.co/Qwen/Qwen3.5-397B-A17B-GPTQ-Int4", "precision": "gptq-int4", "format": "safetensors", "sizeGB": 235.8, "memoryGB": 512 }, { "name": "lmstudio-community Qwen3.5-397B-A17B-MLX-4bit", "url": "https://huggingface.co/lmstudio-community/Qwen3.5-397B-A17B-MLX-4bit", "precision": "mlx-4bit", "format": "mlx", "sizeGB": 223.9, "memoryGB": 512, "runsOn": [ "LM Studio (Apple silicon)" ] }, { "name": "lmstudio-community Qwen3.5-397B-A17B-GGUF Q4_K_M", "url": "https://huggingface.co/lmstudio-community/Qwen3.5-397B-A17B-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 241.5, "memoryGB": 512, "runsOn": [ "llama.cpp", "LM Studio" ] }, { "name": "Qwen3.5-397B-A17B", "url": "https://huggingface.co/Qwen/Qwen3.5-397B-A17B", "precision": "bf16", "format": "safetensors", "sizeGB": 806.8, "memoryGB": 1024, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "gemma-4-e4b-it", "name": "Gemma 4 E4B IT", "grade": "B", "score": 63.2, "tasks": [ "agents", "general", "vision" ], "paramsB": 8, "params": "8B", "context": 131072, "goodFor": "Small on-device assistant and agent model, 4.5B effective, with tool calling and optional thinking, taking text, images, video frames and short audio clips for transcription and translation.", "licence": "Apache-2.0", "run": "llama-server -hf google/gemma-4-E4B-it-qat-q4_0-gguf --hf-file gemma-4-E4B_q4_0-it.gguf --jinja -c 131072", "runBuild": "gemma-4-E4B_q4_0-it", "url": "/tools/gemma-4-e4b-it", "builds": [ { "name": "gemma-4-E4B_q4_0-it", "url": "https://huggingface.co/google/gemma-4-E4B-it-qat-q4_0-gguf/blob/main/gemma-4-E4B_q4_0-it.gguf", "precision": "q4_0", "format": "gguf", "sizeGB": 6.15, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)", "LM Studio" ] }, { "name": "gemma-4-E4B-it-qat-mobile-ct", "url": "https://huggingface.co/google/gemma-4-E4B-it-qat-mobile-ct", "precision": "wNa8o8", "format": "safetensors", "sizeGB": 3.77, "memoryGB": 8 }, { "name": "gemma-4-E4B-it-qat-mobile-transformers", "url": "https://huggingface.co/google/gemma-4-E4B-it-qat-mobile-transformers", "precision": "wNa8o8", "format": "safetensors", "sizeGB": 3.56, "memoryGB": 8 }, { "name": "ollama gemma4:e4b", "url": "https://ollama.com/library/gemma4:e4b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 6.6, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "lmstudio google/gemma-4-e4b", "url": "https://lmstudio.ai/models/google/gemma-4-e4b", "precision": "q4_k_m", "format": "gguf", "sizeGB": 5.9, "memoryGB": 8, "runsOn": [ "LM Studio" ] }, { "name": "gemma-4-E4B-it-qat-w4a16-ct", "url": "https://huggingface.co/google/gemma-4-E4B-it-qat-w4a16-ct", "precision": "w4a16", "format": "safetensors", "sizeGB": 11.55, "memoryGB": 16 }, { "name": "gemma-4-E4B-it", "url": "https://huggingface.co/google/gemma-4-E4B-it", "precision": "bf16", "format": "safetensors", "sizeGB": 16.02, "memoryGB": 24, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "qwen3-5-122b-a10b", "name": "Qwen3.5-122B-A10B", "grade": "B", "score": 63.1, "tasks": [ "agents", "general", "vision" ], "params": "122B total, 10B active", "context": 262144, "goodFor": "Mid-size mixture-of-experts vision-language model for coding, agents and tool calling, text, images and video in, thinking on by default.", "licence": "Apache-2.0", "run": "ollama run qwen3.5:122b", "runBuild": "ollama qwen3.5:122b", "url": "/tools/qwen3-5-122b-a10b", "builds": [ { "name": "Qwen3.5-122B-A10B-GPTQ-Int4", "url": "https://huggingface.co/Qwen/Qwen3.5-122B-A10B-GPTQ-Int4", "precision": "gptq-int4", "format": "safetensors", "sizeGB": 78.9, "memoryGB": 96 }, { "name": "lmstudio-community Qwen3.5-122B-A10B-GGUF Q4_K_M", "url": "https://huggingface.co/lmstudio-community/Qwen3.5-122B-A10B-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 75.1, "memoryGB": 96, "runsOn": [ "llama.cpp", "LM Studio" ] }, { "name": "ollama qwen3.5:122b", "url": "https://ollama.com/library/qwen3.5:122b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 81, "memoryGB": 128, "runsOn": [ "Ollama" ] }, { "name": "Qwen3.5-122B-A10B-FP8", "url": "https://huggingface.co/Qwen/Qwen3.5-122B-A10B-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 127.2, "memoryGB": 192 }, { "name": "Qwen3.5-122B-A10B", "url": "https://huggingface.co/Qwen/Qwen3.5-122B-A10B", "precision": "bf16", "format": "safetensors", "sizeGB": 250.2, "memoryGB": 512, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "glm-4-7-flash", "name": "GLM-4.7-Flash", "grade": "B", "score": 63, "tasks": [ "agents", "general" ], "paramsB": 30, "activeB": 3, "params": "30B", "context": 202752, "goodFor": "Small mixture-of-experts reasoning model for English and Chinese with tool calling, a local coding and agent model.", "licence": "MIT", "run": "vllm serve zai-org/GLM-4.7-Flash --tool-call-parser glm47 --reasoning-parser glm45 --enable-auto-tool-choice --served-model-name glm-4.7-flash", "runBuild": "GLM-4.7-Flash", "url": "/tools/glm-4-7-flash", "builds": [ { "name": "ollama glm-4.7-flash:latest", "url": "https://ollama.com/library/glm-4.7-flash:latest", "precision": "q4_k_m", "format": "ollama", "sizeGB": 19, "memoryGB": 24, "runsOn": [ "Ollama" ] }, { "name": "GLM-4.7-Flash", "url": "https://huggingface.co/zai-org/GLM-4.7-Flash", "precision": "bf16", "format": "safetensors", "sizeGB": 62.4, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "qwen3-6-35b-a3b", "name": "Qwen3.6-35B-A3B", "grade": "B", "score": 62.7, "tasks": [ "agents", "general", "vision" ], "paramsB": 35, "activeB": 3, "params": "35B", "context": 262144, "goodFor": "Fast mixture-of-experts vision-language model for coding agents and tool calling, text, images and video in, thinking on by default.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3.6-35B-A3B-FP8 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "Qwen3.6-35B-A3B-FP8", "url": "/tools/qwen3-6-35b-a3b", "builds": [ { "name": "ollama qwen3.6:35b", "url": "https://ollama.com/library/qwen3.6:35b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 23, "memoryGB": 32, "runsOn": [ "Ollama" ] }, { "name": "Qwen3.6-35B-A3B-FP8", "url": "https://huggingface.co/Qwen/Qwen3.6-35B-A3B-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 37.5, "memoryGB": 48 }, { "name": "Qwen3.6-35B-A3B", "url": "https://huggingface.co/Qwen/Qwen3.6-35B-A3B", "precision": "bf16", "format": "safetensors", "sizeGB": 71.9, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "hy3", "name": "Hy3", "grade": "B", "score": 62.6, "tasks": [ "agents", "general" ], "paramsB": 295, "activeB": 21, "params": "295B", "context": 262144, "goodFor": "Mixture-of-experts language model with tool calling and per-request reasoning effort (off by default), text only.", "licence": "Apache-2.0", "run": "vllm serve tencent/Hy3-FP8 --tensor-parallel-size 8 --tool-call-parser hy_v3 --reasoning-parser hy_v3 --enable-auto-tool-choice --served-model-name hy3", "runBuild": "Hy3-FP8", "url": "/tools/hy3", "builds": [ { "name": "Hy3-Q2_K_XL", "url": "https://huggingface.co/AngelSlim/Hy3-GGUF/blob/main/Hy3-Q2_K_XL.gguf", "precision": "q2_k_xl", "format": "gguf", "sizeGB": 100.1, "memoryGB": 128, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Hy3-IQ1_M", "url": "https://huggingface.co/AngelSlim/Hy3-GGUF/blob/main/Hy3-IQ1_M.gguf", "precision": "iq1_m", "format": "gguf", "sizeGB": 89.4, "memoryGB": 128, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Hy3-Q3_K_M", "url": "https://huggingface.co/AngelSlim/Hy3-GGUF/blob/main/Hy3-Q3_K_M.gguf", "precision": "q3_k_m", "format": "gguf", "sizeGB": 134.6, "memoryGB": 192, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Hy3-Q4_K_M", "url": "https://huggingface.co/AngelSlim/Hy3-GGUF/blob/main/Hy3-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 182.3, "memoryGB": 256, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Hy3-GPTQ-Int4", "url": "https://huggingface.co/AngelSlim/Hy3-GPTQ-Int4", "precision": "gptq-int4", "format": "safetensors", "sizeGB": 168.5, "memoryGB": 256 }, { "name": "Hy3-FP8", "url": "https://huggingface.co/tencent/Hy3-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 299.9, "memoryGB": 512 }, { "name": "Hy3", "url": "https://huggingface.co/tencent/Hy3", "precision": "bf16", "format": "safetensors", "sizeGB": 597.6, "memoryGB": 1024, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "qwen3-embedding-0-6b", "name": "Qwen3-Embedding-0.6B", "grade": "B", "score": 62.5, "tasks": [ "embed" ], "paramsB": 0.6, "params": "0.6B", "context": 32768, "goodFor": "Multilingual and code text embeddings for retrieval, classification and clustering, with task instructions on queries and vectors up to 1,024 dimensions.", "licence": "Apache-2.0", "run": "ollama pull qwen3-embedding:0.6b", "runBuild": "ollama qwen3-embedding:0.6b", "url": "/tools/qwen3-embedding-0-6b", "builds": [ { "name": "Qwen3-Embedding-0.6B", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-0.6B", "precision": "bf16", "format": "safetensors", "sizeGB": 1.21, "memoryGB": 8, "runsOn": [ "vLLM", "SGLang", "Transformers" ] }, { "name": "Qwen3-Embedding-0.6B-GGUF Q8_0", "url": "https://huggingface.co/Qwen/Qwen3-Embedding-0.6B-GGUF", "precision": "q8_0", "format": "gguf", "sizeGB": 0.64, "memoryGB": 8, "runsOn": [ "llama.cpp" ] }, { "name": "ollama qwen3-embedding:0.6b", "url": "https://ollama.com/library/qwen3-embedding:0.6b", "precision": "q8_0", "format": "ollama", "sizeGB": 0.64, "memoryGB": 8, "runsOn": [ "Ollama" ] } ] }, { "slug": "kimi-k3", "name": "Kimi K3", "grade": "B", "score": 62.4, "tasks": [ "agents", "general", "vision" ], "paramsB": 2800, "activeB": 104, "params": "2.8T", "context": 1048576, "goodFor": "Large multimodal agent model for coding and long tool-calling sessions, text and images in, with always-on reasoning and JSON Schema output.", "licence": "Kimi K3 License (modified MIT)", "run": "sglang serve --trust-remote-code --model-path moonshotai/Kimi-K3 --tp-size 8 --context-length 65536 --mem-fraction-static 0.82 --reasoning-parser kimi_k3 --tool-call-parser kimi_k3 --port 30000", "runBuild": "Kimi-K3", "url": "/tools/kimi-k3", "builds": [ { "name": "Kimi-K3", "url": "https://huggingface.co/moonshotai/Kimi-K3", "precision": "mxfp4", "format": "safetensors", "sizeGB": 1560.9, "memoryGB": 2048 } ] }, { "slug": "qwen3-vl-embedding-8b", "name": "Qwen3-VL-Embedding-8B", "grade": "B", "score": 62.1, "tasks": [ "embed" ], "paramsB": 8.14, "params": "8.14B", "context": 32768, "goodFor": "Multimodal retrieval and clustering over text, images, screenshots, documents and video in one vector space, with task instructions and shortened vectors.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3-VL-Embedding-8B --runner pooling --max-model-len 32768", "runBuild": "Qwen3-VL-Embedding-8B", "url": "/tools/qwen3-vl-embedding-8b", "builds": [ { "name": "Qwen3-VL-Embedding-8B", "url": "https://huggingface.co/Qwen/Qwen3-VL-Embedding-8B", "precision": "bf16", "format": "safetensors", "sizeGB": 16.3, "memoryGB": 24 } ] }, { "slug": "voxtral-mini-4b-realtime-2602", "name": "Voxtral Mini 4B Realtime 2602", "grade": "B", "score": 62, "tasks": [ "stt" ], "paramsB": 4.43, "params": "4.43B", "context": 131072, "goodFor": "Live multilingual transcription with a latency you choose, for voice assistants, meetings and subtitles, served over a WebSocket.", "licence": "Apache-2.0", "run": "VLLM_DISABLE_COMPILE_CACHE=1 vllm serve mistralai/Voxtral-Mini-4B-Realtime-2602 --compilation_config '{\"cudagraph_mode\": \"PIECEWISE\"}'", "runBuild": "Voxtral-Mini-4B-Realtime-2602", "url": "/tools/voxtral-mini-4b-realtime-2602", "builds": [ { "name": "Voxtral-Mini-4B-Realtime-2602-ExecuTorch (model-metal-int4.pte)", "url": "https://huggingface.co/mistral-experimental/Voxtral-Mini-4B-Realtime-2602-ExecuTorch", "precision": "int4", "format": "executorch", "sizeGB": 4.44, "memoryGB": 8 }, { "name": "AudioCPP-Voxtral-Mini-4B-Realtime-2602-GGUF q4_k", "url": "https://huggingface.co/mistral-experimental/AudioCPP-Voxtral-Mini-4B-Realtime-2602-GGUF", "precision": "q4_k", "format": "gguf", "sizeGB": 3.1, "memoryGB": 8 }, { "name": "Voxtral-Mini-4B-Realtime-2602", "url": "https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602", "precision": "bf16", "format": "safetensors", "sizeGB": 8.86, "memoryGB": 12, "runsOn": [ "vLLM", "Transformers" ] } ] }, { "slug": "deepseek-v4-flash-0731", "name": "DeepSeek-V4-Flash-0731", "grade": "C", "score": 61.9, "tasks": [ "agents", "general" ], "paramsB": 284, "activeB": 13, "params": "284B", "context": 1048576, "goodFor": "Text model for agents and long documents, with thinking or chat mode, reasoning effort and tool calls in DSML tags.", "licence": "MIT", "run": "vllm serve deepseek-ai/DeepSeek-V4-Flash-0731 --trust-remote-code --kv-cache-dtype fp8 --block-size 256 --data-parallel-size 4 --enable-expert-parallel --moe-backend deep_gemm_mega_moe --attention-config '{\"use_fp4_indexer_cache\": true}' --speculative-config '{\"method\":\"dspark\",\"num_speculative_tokens\":7,\"draft_sample_method\":\"greedy\"}' --tokenizer-mode deepseek_v4 --tool-call-parser deepseek_v4 --enable-auto-tool-choice --reasoning-parser deepseek_v4", "runBuild": "DeepSeek-V4-Flash-0731", "url": "/tools/deepseek-v4-flash-0731", "builds": [ { "name": "DeepSeek-V4-Flash-0731", "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731", "precision": "fp8+fp4", "format": "safetensors", "sizeGB": 166.9, "memoryGB": 256 } ] }, { "slug": "qwen3-8-flash-next", "name": "Qwen3.8-Flash-Next", "grade": "C", "score": 61.8, "tasks": [ "agents", "general", "vision" ], "paramsB": 125, "activeB": 6, "params": "125B", "context": 262144, "goodFor": "Experimental preview of the Qwen4 architecture: a multimodal mixture-of-experts model with tool calling and adjustable reasoning effort.", "licence": "Qwen Community License 1.0 (custom, not an OSI licence). Free of charge, the model name shown in products above 100M monthly users or US$20M monthly revenue, and a separate licence from Qwen before commercial use by a Model as a Service or AI Work Assistant business", "run": "vllm serve Qwen/Qwen3.8-Flash-Next-FP8 --tensor-parallel-size 4 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "Qwen3.8-Flash-Next-FP8", "url": "/tools/qwen3-8-flash-next", "builds": [ { "name": "Qwen3.8-Flash-Next-FP4-Local", "url": "https://huggingface.co/nvidia/Qwen3.8-Flash-Next-FP4-Local", "precision": "nvfp4", "format": "safetensors", "sizeGB": 102.2, "memoryGB": 128 }, { "name": "ollama qwen3.8-flash-next:125b-a6b-q4_K_M", "url": "https://ollama.com/library/qwen3.8-flash-next:125b-a6b-q4_K_M", "precision": "q4_k_m", "format": "ollama", "sizeGB": 120, "memoryGB": 192, "runsOn": [ "Ollama" ] }, { "name": "Qwen3.8-Flash-Next-NVFP4", "url": "https://huggingface.co/nvidia/Qwen3.8-Flash-Next-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 132.7, "memoryGB": 192 }, { "name": "Qwen3.8-Flash-Next-FP8", "url": "https://huggingface.co/Qwen/Qwen3.8-Flash-Next-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 185.5, "memoryGB": 256 }, { "name": "Qwen3.8-Flash-Next", "url": "https://huggingface.co/Qwen/Qwen3.8-Flash-Next", "precision": "bf16", "format": "safetensors", "sizeGB": 360, "memoryGB": 512, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "deepseek-v4-1-flash", "name": "DeepSeek V4.1 Flash", "grade": "C", "score": 61.5, "tasks": [ "agents", "general", "vision" ], "paramsB": 552, "params": "552B", "context": 1048576, "goodFor": "Multimodal mixture-of-experts model for agents and long context, text and images in, with tool calls and a reasoning budget.", "licence": "MIT", "run": "pip install -r inference/requirements.txt \u0026\u0026 python convert.py --hf-ckpt-path \"$HF_CKPT_PATH\" --save-path \"$SAVE_PATH\" --model-parallel 8 --expert-dtype fp4 --tokenizer-path \"$HF_CKPT_PATH\" \u0026\u0026 torchrun --nproc-per-node 8 generate.py --ckpt-path \"$SAVE_PATH\" --config config.json --interactive", "runBuild": "DeepSeek-V4.1-Flash-1017", "url": "/tools/deepseek-v4-1-flash", "builds": [ { "name": "DeepSeek-V4.1-Flash-1017", "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017", "precision": "fp8+fp4", "format": "safetensors", "sizeGB": 510.3, "memoryGB": 1024 } ] }, { "slug": "nemotron-3-nano-omni-30b-a3b-reasoning", "name": "Nemotron 3 Nano Omni 30B A3B Reasoning", "grade": "C", "score": 61.2, "tasks": [ "agents", "general", "vision" ], "paramsB": 31, "activeB": 3, "params": "31B", "context": 262144, "goodFor": "Multimodal understanding of video with speech, documents, charts and screens, and English transcription, with tool calling and a reasoning budget.", "licence": "NVIDIA Open Model Agreement (custom, commercial use allowed, not an OSI licence)", "run": "vllm serve nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8 --max-model-len 131072 --trust-remote-code --video-pruning-rate 0.5 --media-io-kwargs '{\"video\": {\"fps\": 2, \"num_frames\": 256}}' --reasoning-parser nemotron_v3 --enable-auto-tool-choice --tool-call-parser qwen3_coder --kv-cache-dtype fp8", "runBuild": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8", "url": "/tools/nemotron-3-nano-omni-30b-a3b-reasoning", "builds": [ { "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4", "url": "https://huggingface.co/nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 22.4, "memoryGB": 32 }, { "name": "ollama nemotron3:33b-nvfp4", "url": "https://ollama.com/library/nemotron3:33b-nvfp4", "precision": "nvfp4", "format": "ollama", "sizeGB": 21, "memoryGB": 32, "runsOn": [ "Ollama" ] }, { "name": "LM Studio nvidia/nemotron-3-nano-omni Q4_K_M", "url": "https://huggingface.co/lmstudio-community/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 26.1, "memoryGB": 32, "runsOn": [ "LM Studio", "llama.cpp" ] }, { "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8", "url": "https://huggingface.co/nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 35.2, "memoryGB": 48 }, { "name": "ollama nemotron3:33b-q4_K_M", "url": "https://ollama.com/library/nemotron3:33b-q4_K_M", "precision": "q4_k_m", "format": "ollama", "sizeGB": 28, "memoryGB": 48, "runsOn": [ "Ollama" ] }, { "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16", "url": "https://huggingface.co/nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16", "precision": "bf16", "format": "safetensors", "sizeGB": 66.1, "memoryGB": 96 } ] }, { "slug": "qwen3-asr-0-6b", "name": "Qwen3-ASR-0.6B", "grade": "C", "score": 61.2, "tasks": [ "stt" ], "paramsB": 0.78, "params": "0.78B", "context": 65536, "goodFor": "Fast, small multilingual and Chinese-dialect transcription with language identification, context hints and realtime streaming through vLLM.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3-ASR-0.6B", "runBuild": "Qwen3-ASR-0.6B", "url": "/tools/qwen3-asr-0-6b", "builds": [ { "name": "Qwen3-ASR-0.6B", "url": "https://huggingface.co/Qwen/Qwen3-ASR-0.6B", "precision": "bf16", "format": "safetensors", "sizeGB": 1.88, "memoryGB": 8, "runsOn": [ "vLLM", "SGLang" ] }, { "name": "Qwen3-ASR-0.6B-hf", "url": "https://huggingface.co/Qwen/Qwen3-ASR-0.6B-hf", "precision": "bf16", "format": "safetensors", "sizeGB": 1.58, "memoryGB": 8, "runsOn": [ "Transformers" ] }, { "name": "ggml-org Qwen3-ASR-0.6B-GGUF Q8_0", "url": "https://huggingface.co/ggml-org/Qwen3-ASR-0.6B-GGUF", "precision": "q8_0", "format": "gguf", "sizeGB": 1.02, "memoryGB": 8, "runsOn": [ "llama.cpp" ] } ] }, { "slug": "qwen3-5-4b", "name": "Qwen3.5-4B", "grade": "C", "score": 61.2, "tasks": [ "agents", "general", "vision" ], "paramsB": 4, "params": "4B", "context": 262144, "goodFor": "Small vision-language model for local use: text, images and video in, tool calling, thinking on by default.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3.5-4B --port 8000 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "Qwen3.5-4B", "url": "/tools/qwen3-5-4b", "builds": [ { "name": "ollama qwen3.5:4b", "url": "https://ollama.com/library/qwen3.5:4b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 3.3, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "Qwen3.5-4B", "url": "https://huggingface.co/Qwen/Qwen3.5-4B", "precision": "bf16", "format": "safetensors", "sizeGB": 9.3, "memoryGB": 12, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "agents-a1-4b", "name": "Agents-A1-4B", "grade": "C", "score": 61, "tasks": [ "agents", "general", "vision" ], "paramsB": 4.5, "params": "4.5B", "context": 262144, "goodFor": "Small dense vision-language model trained for long-horizon agent work, search, research, instruction following and tool calling, thinking on by default, runs on almost any machine.", "licence": "Apache-2.0", "run": "vllm serve InternScience/Agents-A1-4B --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "Agents-A1-4B", "url": "/tools/agents-a1-4b", "builds": [ { "name": "Agents-A1-4B-Q8_0", "url": "https://huggingface.co/InternScience/Agents-A1-4B-Q8_0-GGUF", "precision": "q8_0", "format": "gguf", "sizeGB": 5.2, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Agents-A1-4B-Q4_K_M", "url": "https://huggingface.co/InternScience/Agents-A1-4B-Q4_K_M-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 3.4, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Agents-A1-4B", "url": "https://huggingface.co/InternScience/Agents-A1-4B", "precision": "bf16", "format": "safetensors", "sizeGB": 9.1, "memoryGB": 12, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)", "Transformers" ] }, { "name": "Agents-A1-4B-FP8", "url": "https://huggingface.co/InternScience/Agents-A1-4B-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 7.8, "memoryGB": 12, "runsOn": [ "vLLM", "SGLang" ] }, { "name": "Agents-A1-4B-F16", "url": "https://huggingface.co/InternScience/Agents-A1-4B-F16-GGUF", "precision": "f16", "format": "gguf", "sizeGB": 9.1, "memoryGB": 12, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] } ] }, { "slug": "apertus-v1-5-8b", "name": "Apertus v1.5 8B", "grade": "C", "score": 61, "tasks": [ "agents", "general", "vision" ], "paramsB": 8.9, "params": "8.9B", "context": 262144, "goodFor": "Fully open multilingual model for text, images and audio in with text out, instruction following, tool calls and optional thinking, for teams that want open data and a European publisher.", "licence": "Apache-2.0, with the Apertus Acceptable Use Policy (v1.5, 14 July 2026) accepted at the download gate", "run": "docker pull ghcr.io/swiss-ai/vllm_apertus_1.5_release:latest-amd64 \u0026\u0026 vllm serve swiss-ai/Apertus-v1.5-8B --chat-template-content-format string --gpu-memory-utilization 0.6 --max-model-len 262144 --enable-auto-tool-choice --tool-call-parser apertus", "runBuild": "Apertus-v1.5-8B", "url": "/tools/apertus-v1-5-8b", "builds": [ { "name": "Apertus-v1.5-8B", "url": "https://huggingface.co/swiss-ai/Apertus-v1.5-8B", "precision": "bf16", "format": "safetensors", "sizeGB": 18.4, "memoryGB": 24, "runsOn": [ "vLLM (Swiss AI fork)", "Transformers (Swiss AI fork)" ] } ] }, { "slug": "apodex-1-1-mini", "name": "Apodex-1.1-mini", "grade": "C", "score": 61, "tasks": [ "agents", "general", "vision" ], "activeB": 3, "params": "35B (3B active)", "context": 262144, "goodFor": "Mixture-of-experts agent model for long-horizon work with files, data, code and tools, fine-tuned from Qwen3.5-35B-A3B, with documented tool calling, parallel calls and thinking on by default.", "licence": "Apache-2.0", "run": "vllm serve apodex/Apodex-1.1-mini --max-model-len 262144 --enable-auto-tool-choice --tool-call-parser qwen3_coder --reasoning-parser qwen3", "runBuild": "Apodex-1.1-mini", "url": "/tools/apodex-1-1-mini", "builds": [ { "name": "Apodex-1.1-mini-NVFP4", "url": "https://huggingface.co/apodex/Apodex-1.1-mini-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 24.3, "memoryGB": 32 }, { "name": "Apodex-1.1-mini-GPTQ-Int4", "url": "https://huggingface.co/apodex/Apodex-1.1-mini-GPTQ-Int4", "precision": "int4", "format": "safetensors", "sizeGB": 24.7, "memoryGB": 32, "runsOn": [ "SGLang" ] }, { "name": "apodex1.1-35b-Q4_K_M-MTP", "url": "https://huggingface.co/apodex/Apodex-1.1-mini-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 22.6, "memoryGB": 32, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Apodex-1.1-mini-FP8", "url": "https://huggingface.co/apodex/Apodex-1.1-mini-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 37.5, "memoryGB": 48, "runsOn": [ "vLLM", "SGLang" ] }, { "name": "Apodex-1.1-mini", "url": "https://huggingface.co/apodex/Apodex-1.1-mini", "precision": "bf16", "format": "safetensors", "sizeGB": 71.9, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "carbon-x", "name": "Carbon-X", "grade": "C", "score": 61, "tasks": [ "general" ], "paramsB": 14.8, "params": "14.8B", "context": 40960, "goodFor": "A DNA specialist another agent calls, for variant-effect questions, regulatory regions and next-base prediction on raw DNA kept on local hardware.", "licence": "Apache-2.0", "run": "llama-server -hf HuggingFaceBio/Carbon-X-GGUF:Q8_0 -ngl 99 -c 40960 --jinja --chat-template-kwargs '{\"enable_thinking\": false}' --temp 0.6 --top-p 0.95 --top-k 20 --min-p 0 --port 8080", "runBuild": "Carbon-X-GGUF Q8_0", "url": "/tools/carbon-x", "builds": [ { "name": "Carbon-X-GGUF Q4_K_M", "url": "https://huggingface.co/HuggingFaceBio/Carbon-X-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 9.64, "memoryGB": 12, "runsOn": [ "llama.cpp" ] }, { "name": "Carbon-X-GGUF Q6_K", "url": "https://huggingface.co/HuggingFaceBio/Carbon-X-GGUF", "precision": "q6_k", "format": "gguf", "sizeGB": 12.56, "memoryGB": 16, "runsOn": [ "llama.cpp" ] }, { "name": "Carbon-X-GGUF Q5_K_M", "url": "https://huggingface.co/HuggingFaceBio/Carbon-X-GGUF", "precision": "q5_k_m", "format": "gguf", "sizeGB": 11.05, "memoryGB": 16, "runsOn": [ "llama.cpp" ] }, { "name": "Carbon-X-GGUF Q8_0", "url": "https://huggingface.co/HuggingFaceBio/Carbon-X-GGUF", "precision": "q8_0", "format": "gguf", "sizeGB": 15.76, "memoryGB": 24, "runsOn": [ "llama.cpp" ] }, { "name": "Carbon-X", "url": "https://huggingface.co/HuggingFaceBio/Carbon-X", "precision": "bf16", "format": "safetensors", "sizeGB": 29.64, "memoryGB": 48, "runsOn": [ "vLLM", "Transformers" ] }, { "name": "Carbon-X-GGUF BF16", "url": "https://huggingface.co/HuggingFaceBio/Carbon-X-GGUF", "precision": "bf16", "format": "gguf", "sizeGB": 29.65, "memoryGB": 48, "runsOn": [ "llama.cpp" ] } ] }, { "slug": "darwin-36b-opus", "name": "Darwin-36B-Opus", "grade": "C", "score": 61, "tasks": [ "general" ], "activeB": 3, "params": "35B (3B active)", "context": 262144, "goodFor": "Text-only mixture-of-experts reasoning model merged from Qwen3.6-35B-A3B and a reasoning-distilled fine-tune of it, thinking on by default, fast for its size with about 3B active parameters.", "licence": "Apache-2.0", "run": "vllm serve FINAL-Bench/Darwin-36B-Opus --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "Darwin-36B-Opus", "url": "/tools/darwin-36b-opus", "builds": [ { "name": "POCKET-35B-IQ1_M", "url": "https://huggingface.co/FINAL-Bench/POCKET-35B-GGUF/blob/main/POCKET-35B-IQ1_M.gguf", "precision": "iq1_m", "format": "gguf", "sizeGB": 8.2, "memoryGB": 12, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "POCKET-35B-Q2_K", "url": "https://huggingface.co/FINAL-Bench/POCKET-35B-GGUF/blob/main/POCKET-35B-Q2_K.gguf", "precision": "q2_k", "format": "gguf", "sizeGB": 12.9, "memoryGB": 16, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "POCKET-35B-Q3_K_M", "url": "https://huggingface.co/FINAL-Bench/POCKET-35B-GGUF/blob/main/POCKET-35B-Q3_K_M.gguf", "precision": "q3_k_m", "format": "gguf", "sizeGB": 16.8, "memoryGB": 24, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "POCKET-35B-Q4_K_M", "url": "https://huggingface.co/FINAL-Bench/POCKET-35B-GGUF/blob/main/POCKET-35B-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 21.2, "memoryGB": 32, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Darwin-36B-Opus", "url": "https://huggingface.co/FINAL-Bench/Darwin-36B-Opus", "precision": "bf16", "format": "safetensors", "sizeGB": 69.4, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)", "Transformers" ] } ] }, { "slug": "kolibri-1", "name": "Kolibri-1", "grade": "C", "score": 61, "tasks": [ "agents", "general" ], "paramsB": 78.1, "activeB": 3.46, "params": "78.1B", "context": 262144, "goodFor": "German and English reasoning model with tool calling and adjustable reasoning effort, text only.", "licence": "Apache-2.0", "run": "pip install 'aleph-alpha-inference\u003e=1' \u0026\u0026 vllm serve Aleph-Alpha/Kolibri-1 --kv-cache-dtype fp8 --reasoning-parser kolibri1 --tool-call-parser kolibri1 --enable-auto-tool-choice", "runBuild": "Kolibri-1", "url": "/tools/kolibri-1", "builds": [ { "name": "Kolibri-1", "url": "https://huggingface.co/Aleph-Alpha/Kolibri-1", "precision": "fp8", "format": "safetensors", "sizeGB": 78.8, "memoryGB": 96 }, { "name": "Kolibri-1-BF16", "url": "https://huggingface.co/Aleph-Alpha/Kolibri-1-BF16", "precision": "bf16", "format": "safetensors", "sizeGB": 156.2, "memoryGB": 192, "runsOn": [ "vLLM (ROCm)" ] } ] }, { "slug": "laguna-s-2-1", "name": "Laguna S 2.1", "grade": "C", "score": 61, "tasks": [ "agents", "general", "coding" ], "params": "117.6B total, 8B active", "context": 1048576, "goodFor": "Long-horizon agentic coding and terminal work with up to 1M tokens of context, on a 128 GB machine at 4-bit or a multi-GPU server.", "licence": "OpenMDW-1.1", "run": "llama-server -hf poolside/Laguna-S-2.1-GGUF:Q4_K_M --jinja -c 65536 --port 8000", "runBuild": "Laguna-S-2.1-GGUF Q4_K_M", "url": "/tools/laguna-s-2-1", "builds": [ { "name": "Laguna-S-2.1-NVFP4-mlx", "url": "https://huggingface.co/poolside/Laguna-S-2.1-NVFP4-mlx", "precision": "nvfp4", "format": "mlx", "sizeGB": 71.9, "memoryGB": 96, "runsOn": [ "MLX" ] }, { "name": "ollama laguna-s-2.1:nvfp4", "url": "https://ollama.com/library/laguna-s-2.1:nvfp4", "precision": "nvfp4", "format": "mlx", "sizeGB": 66, "memoryGB": 96, "runsOn": [ "Ollama" ] }, { "name": "Laguna-S-2.1-INT4", "url": "https://huggingface.co/poolside/Laguna-S-2.1-INT4", "precision": "int4 (w4a16)", "format": "safetensors", "sizeGB": 99.7, "memoryGB": 128 }, { "name": "Laguna-S-2.1-NVFP4", "url": "https://huggingface.co/poolside/Laguna-S-2.1-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 99.7, "memoryGB": 128 }, { "name": "Laguna-S-2.1-GGUF Q4_K_M", "url": "https://huggingface.co/poolside/Laguna-S-2.1-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 96, "memoryGB": 128, "runsOn": [ "llama.cpp" ] }, { "name": "ollama laguna-s-2.1:latest", "url": "https://ollama.com/library/laguna-s-2.1:latest", "precision": "q4_k_m", "format": "ollama", "sizeGB": 96, "memoryGB": 128, "runsOn": [ "Ollama" ] }, { "name": "Laguna-S-2.1-FP8", "url": "https://huggingface.co/poolside/Laguna-S-2.1-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 131.3, "memoryGB": 192 }, { "name": "Laguna-S-2.1", "url": "https://huggingface.co/poolside/Laguna-S-2.1", "precision": "bf16", "format": "safetensors", "sizeGB": 235.1, "memoryGB": 512, "runsOn": [ "vLLM", "SGLang", "Transformers" ] } ] }, { "slug": "laguna-xs-2-1", "name": "Laguna XS 2.1", "grade": "C", "score": 61, "tasks": [ "agents", "general", "coding" ], "params": "33.4B total, 3B active", "context": 262144, "goodFor": "Agentic coding and long-horizon terminal work on one GPU or a 32 GB machine, with tool calling and reasoning that turns on or off per request.", "licence": "OpenMDW-1.1", "run": "llama-server -hf poolside/Laguna-XS-2.1-GGUF:Q4_K_M --jinja -c 131072 --port 8000", "runBuild": "Laguna-XS-2.1-GGUF Q4_K_M", "url": "/tools/laguna-xs-2-1", "builds": [ { "name": "Laguna-XS-2.1-INT4", "url": "https://huggingface.co/poolside/Laguna-XS-2.1-INT4", "precision": "int4 (w4a16)", "format": "safetensors", "sizeGB": 23.9, "memoryGB": 32 }, { "name": "Laguna-XS-2.1-NVFP4", "url": "https://huggingface.co/poolside/Laguna-XS-2.1-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 21.6, "memoryGB": 32 }, { "name": "Laguna-XS-2.1-NVFP4-mlx", "url": "https://huggingface.co/poolside/Laguna-XS-2.1-NVFP4-mlx", "precision": "nvfp4", "format": "mlx", "sizeGB": 21.6, "memoryGB": 32, "runsOn": [ "MLX" ] }, { "name": "Laguna-XS-2.1-GGUF Q4_K_M", "url": "https://huggingface.co/poolside/Laguna-XS-2.1-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 20.3, "memoryGB": 32, "runsOn": [ "llama.cpp", "Ollama" ] }, { "name": "ollama laguna-xs-2.1:latest", "url": "https://ollama.com/library/laguna-xs-2.1:latest", "precision": "q4_k_m", "format": "ollama", "sizeGB": 20.3, "memoryGB": 32, "runsOn": [ "Ollama" ] }, { "name": "Laguna-XS-2.1-FP8", "url": "https://huggingface.co/poolside/Laguna-XS-2.1-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 35.3, "memoryGB": 48 }, { "name": "Laguna-XS-2.1", "url": "https://huggingface.co/poolside/Laguna-XS-2.1", "precision": "bf16", "format": "safetensors", "sizeGB": 66.9, "memoryGB": 96, "runsOn": [ "vLLM", "SGLang", "Transformers" ] } ] }, { "slug": "mimo-v2-6-distill-qwen-9b", "name": "MiMo-V2.6 Distill Qwen 9B", "grade": "C", "score": 61, "tasks": [ "agents", "general", "vision" ], "paramsB": 9.41, "params": "9.41B", "context": 262144, "goodFor": "A 9B agentic model for coding, terminal and tool tasks, with images in, meant as a starting point for RL fine-tuning.", "licence": "MIT (metadata only, no LICENSE file); built on Qwen3.5-9B under Apache 2.0", "run": "sglang serve --model-path XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B --reasoning-parser mimo --tool-call-parser mimo --context-length 131072 --port 30000", "runBuild": "MiMo-V2.6-Distill-Qwen-9B", "url": "/tools/mimo-v2-6-distill-qwen-9b", "builds": [ { "name": "ollama mimo-v2.6:9b-q4_K_M", "url": "https://ollama.com/library/mimo-v2.6:9b-q4_K_M", "precision": "q4_k_m", "format": "ollama", "sizeGB": 6.6, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "MiMo-V2.6-Distill-Qwen-9B", "url": "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B", "precision": "bf16", "format": "safetensors", "sizeGB": 18.85, "memoryGB": 24, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)", "Transformers" ] } ] }, { "slug": "minicpm-o-4-5", "name": "MiniCPM-o 4.5", "grade": "C", "score": 61, "tasks": [ "general", "vision" ], "paramsB": 9.4, "params": "9.4B", "context": 40960, "goodFor": "Omni-modal assistant for images, video, speech and text in, text and speech out, with live full-duplex conversation and voice cloning, small enough for a laptop at 4-bit.", "licence": "Apache-2.0; built from Qwen3-8B, SigLIP2, Whisper-medium and CosyVoice2, all Apache 2.0 on their Hub cards", "run": "vllm serve openbmb/MiniCPM-o-4_5 --dtype auto --trust-remote-code --max-model-len 4096 --max-num-batched-tokens 2048", "runBuild": "MiniCPM-o-4_5", "url": "/tools/minicpm-o-4-5", "builds": [ { "name": "MiniCPM-o-4_5-gguf Q4_K_M", "url": "https://huggingface.co/openbmb/MiniCPM-o-4_5-gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 6.12, "memoryGB": 8, "runsOn": [ "llama.cpp" ] }, { "name": "MiniCPM-o-4_5-gguf Q4_0", "url": "https://huggingface.co/openbmb/MiniCPM-o-4_5-gguf", "precision": "q4_0", "format": "gguf", "sizeGB": 5.87, "memoryGB": 8 }, { "name": "ollama openbmb/minicpm-o4.5:latest", "url": "https://ollama.com/openbmb/minicpm-o4.5/tags", "precision": "q4_k_m", "format": "ollama", "sizeGB": 6.1, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "MiniCPM-o-4_5-gguf Q8_0", "url": "https://huggingface.co/openbmb/MiniCPM-o-4_5-gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 9.8, "memoryGB": 12 }, { "name": "MiniCPM-o-4_5-GPTQ", "url": "https://huggingface.co/openbmb/MiniCPM-o-4_5-GPTQ", "precision": "gptq-int4", "format": "safetensors", "sizeGB": 9.77, "memoryGB": 12 }, { "name": "MiniCPM-o-4_5-awq", "url": "https://huggingface.co/openbmb/MiniCPM-o-4_5-awq", "precision": "awq-int4", "format": "safetensors", "sizeGB": 13.07, "memoryGB": 16 }, { "name": "MiniCPM-o-4_5-gguf F16", "url": "https://huggingface.co/openbmb/MiniCPM-o-4_5-gguf", "precision": "f16", "format": "gguf", "sizeGB": 17.48, "memoryGB": 24 }, { "name": "MiniCPM-o-4_5", "url": "https://huggingface.co/openbmb/MiniCPM-o-4_5", "precision": "bf16", "format": "safetensors", "sizeGB": 20.05, "memoryGB": 32, "runsOn": [ "vLLM", "Transformers (remote code)" ] } ] }, { "slug": "nemotron-3-5-bolt-e4b-4gb", "name": "NVIDIA Nemotron 3.5 Bolt E4B 4GB", "grade": "C", "score": 61, "tasks": [ "agents", "general" ], "paramsB": 8, "params": "8B", "context": 65536, "goodFor": "A small, fast cousin for voice and device assistants, game characters and robot interfaces that interleave short replies with multi-turn tool calls on a consumer GPU or a Jetson.", "licence": "OpenMDW-1.1", "run": "llama-server -m Nemotron-3.5-Bolt-E4B-Q4_K_M.gguf --alias nemotron-bolt -c 65536 -ngl 999 --jinja --port 8000", "runBuild": "NVIDIA-Nemotron-3.5-Bolt-E4B-4GB (Q4_K_M)", "url": "/tools/nemotron-3-5-bolt-e4b-4gb", "builds": [ { "name": "NVIDIA-Nemotron-3.5-Bolt-E4B-4GB (Q4_K_M)", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB", "precision": "q4_k_m", "format": "gguf", "sizeGB": 6.18, "memoryGB": 8, "runsOn": [ "llama.cpp" ] }, { "name": "lmstudio-community NVIDIA-Nemotron-3.5-Bolt-E4B-MLX-4bit", "url": "https://huggingface.co/lmstudio-community/NVIDIA-Nemotron-3.5-Bolt-E4B-MLX-4bit", "precision": "4bit", "format": "mlx", "sizeGB": 4.23, "memoryGB": 8, "runsOn": [ "LM Studio" ] }, { "name": "lmstudio-community NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF (Q4_K_M)", "url": "https://huggingface.co/lmstudio-community/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 6.18, "memoryGB": 8, "runsOn": [ "LM Studio", "llama.cpp" ] }, { "name": "ollama nemotron-3.5-bolt:e4b-q4_K_M", "url": "https://ollama.com/library/nemotron-3.5-bolt:e4b-q4_K_M", "precision": "q4_k_m", "format": "ollama", "sizeGB": 7.9, "memoryGB": 12, "runsOn": [ "Ollama" ] }, { "name": "ollama nemotron-3.5-bolt:e4b-nvfp4", "url": "https://ollama.com/library/nemotron-3.5-bolt:e4b-nvfp4", "precision": "nvfp4", "format": "ollama", "sizeGB": 7.2, "memoryGB": 12, "runsOn": [ "Ollama (MLX)" ] }, { "name": "ggml-org NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF (Q8_0)", "url": "https://huggingface.co/ggml-org/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF", "precision": "q8_0", "format": "gguf", "sizeGB": 8.03, "memoryGB": 12, "runsOn": [ "llama.cpp" ] }, { "name": "NVIDIA-Nemotron-3.5-Bolt-E4B-BF16", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-BF16", "precision": "bf16", "format": "safetensors", "sizeGB": 16.06, "memoryGB": 24, "runsOn": [ "vLLM", "Transformers", "SGLang" ] } ] }, { "slug": "ornith-1-5-35b-a3b", "name": "Ornith-1.5-35B-A3B", "grade": "C", "score": 61, "tasks": [ "agents", "general", "coding", "vision" ], "params": "35B (3B active)", "context": 262144, "goodFor": "Coding and terminal-agent mixture-of-experts model with tool calling and thinking on by default, text and images in, fast for its size with about 3B active parameters.", "licence": "MIT", "run": "ollama run ornith-1.5:35b", "runBuild": "ollama ornith-1.5:35b", "url": "/tools/ornith-1-5-35b-a3b", "builds": [ { "name": "Ornith-1.5-35B-A3B-MLX-4bit", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B-MLX-4bit", "precision": "4bit", "format": "mlx", "sizeGB": 19.5, "memoryGB": 24 }, { "name": "Ornith-1.5-35B-A3B-NVFP4", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 23.5, "memoryGB": 32 }, { "name": "Ornith-1.5-35B-Q5_K_M", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B-GGUF/blob/main/Ornith-1.5-35B-Q5_K_M.gguf", "precision": "q5_k_m", "format": "gguf", "sizeGB": 25.3, "memoryGB": 32, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Ornith-1.5-35B-Q4_K_M", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B-GGUF/blob/main/Ornith-1.5-35B-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 21.7, "memoryGB": 32, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "ollama ornith-1.5:35b", "url": "https://ollama.com/library/ornith-1.5:35b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 23, "memoryGB": 32, "runsOn": [ "Ollama" ] }, { "name": "Ornith-1.5-35B-A3B-FP8", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 39.4, "memoryGB": 48, "runsOn": [ "vLLM", "SGLang" ] }, { "name": "Ornith-1.5-35B-Q8_0", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B-GGUF/blob/main/Ornith-1.5-35B-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 37.8, "memoryGB": 48, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Ornith-1.5-35B-Q6_K", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B-GGUF/blob/main/Ornith-1.5-35B-Q6_K.gguf", "precision": "q6_k", "format": "gguf", "sizeGB": 29.2, "memoryGB": 48, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Ornith-1.5-35B-A3B-MLX-8bit", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B-MLX-8bit", "precision": "8bit", "format": "mlx", "sizeGB": 36.9, "memoryGB": 48 }, { "name": "Ornith-1.5-35B-A3B-MLX-6bit", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B-MLX-6bit", "precision": "6bit", "format": "mlx", "sizeGB": 28.2, "memoryGB": 48 }, { "name": "Ornith-1.5-35B-A3B", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B", "precision": "bf16", "format": "safetensors", "sizeGB": 71.9, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] }, { "name": "Ornith-1.5-35B-BF16", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B-GGUF/blob/main/Ornith-1.5-35B-BF16.gguf", "precision": "bf16", "format": "gguf", "sizeGB": 71.1, "memoryGB": 96, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Ornith-1.5-35B-A3B-MLX", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-35B-A3B-MLX", "precision": "bf16", "format": "mlx", "sizeGB": 69.3, "memoryGB": 96 } ] }, { "slug": "ornith-1-5-9b", "name": "Ornith-1.5-9B", "grade": "C", "score": 61, "tasks": [ "agents", "general", "coding", "vision" ], "paramsB": 9, "params": "9B", "context": 262144, "goodFor": "Coding and terminal-agent model with tool calling and thinking on by default, text and images in.", "licence": "MIT", "run": "ollama run ornith-1.5:9b", "runBuild": "ollama ornith-1.5:9b", "url": "/tools/ornith-1-5-9b", "builds": [ { "name": "Ornith-1.5-9B-Q5_K_M", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-9B-GGUF/blob/main/Ornith-1.5-9B-Q5_K_M.gguf", "precision": "q5_k_m", "format": "gguf", "sizeGB": 6.6, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Ornith-1.5-9B-Q4_K_M", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-9B-GGUF/blob/main/Ornith-1.5-9B-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 5.8, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Ornith-1.5-9B-MLX-4bit", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-9B-MLX-4bit", "precision": "4bit", "format": "mlx", "sizeGB": 5, "memoryGB": 8 }, { "name": "ollama ornith-1.5:9b", "url": "https://ollama.com/library/ornith-1.5:9b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 6.6, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "Ornith-1.5-9B-Q8_0", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-9B-GGUF/blob/main/Ornith-1.5-9B-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 9.8, "memoryGB": 12, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Ornith-1.5-9B-Q6_K", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-9B-GGUF/blob/main/Ornith-1.5-9B-Q6_K.gguf", "precision": "q6_k", "format": "gguf", "sizeGB": 7.6, "memoryGB": 12, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Ornith-1.5-9B-MLX-8bit", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-9B-MLX-8bit", "precision": "8bit", "format": "mlx", "sizeGB": 9.5, "memoryGB": 12 }, { "name": "Ornith-1.5-9B-MLX-6bit", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-9B-MLX-6bit", "precision": "6bit", "format": "mlx", "sizeGB": 7.3, "memoryGB": 12 }, { "name": "Ornith-1.5-9B-NVFP4", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-9B-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 8.8, "memoryGB": 12 }, { "name": "Ornith-1.5-9B", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-9B", "precision": "bf16", "format": "safetensors", "sizeGB": 19.3, "memoryGB": 24, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] }, { "name": "Ornith-1.5-9B-BF16", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-9B-GGUF/blob/main/Ornith-1.5-9B-BF16.gguf", "precision": "bf16", "format": "gguf", "sizeGB": 18.4, "memoryGB": 24, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Ornith-1.5-9B-MLX", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-9B-MLX", "precision": "bf16", "format": "mlx", "sizeGB": 17.9, "memoryGB": 24 } ] }, { "slug": "qwen3-vl-2b-instruct", "name": "Qwen3-VL-2B-Instruct", "grade": "C", "score": 61, "tasks": [ "agents", "general", "vision" ], "paramsB": 2.13, "params": "2.13B", "context": 262144, "goodFor": "Small vision-language model for images, documents and video with text out, light enough for a laptop or edge device, no thinking mode in this Instruct edition.", "licence": "Apache-2.0", "run": "llama-server -hf Qwen/Qwen3-VL-2B-Instruct-GGUF:Q4_K_M --jinja -c 32768", "runBuild": "Qwen3VL-2B-Instruct-Q4_K_M", "url": "/tools/qwen3-vl-2b-instruct", "builds": [ { "name": "Qwen3-VL-2B-Instruct", "url": "https://huggingface.co/Qwen/Qwen3-VL-2B-Instruct", "precision": "bf16", "format": "safetensors", "sizeGB": 4.27, "memoryGB": 8, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] }, { "name": "Qwen3-VL-2B-Instruct-FP8", "url": "https://huggingface.co/Qwen/Qwen3-VL-2B-Instruct-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 3.48, "memoryGB": 8 }, { "name": "Qwen3VL-2B-Instruct-F16", "url": "https://huggingface.co/Qwen/Qwen3-VL-2B-Instruct-GGUF/blob/main/Qwen3VL-2B-Instruct-F16.gguf", "precision": "f16", "format": "gguf", "sizeGB": 4.27, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Qwen3VL-2B-Instruct-Q8_0", "url": "https://huggingface.co/Qwen/Qwen3-VL-2B-Instruct-GGUF/blob/main/Qwen3VL-2B-Instruct-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 2.28, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Qwen3VL-2B-Instruct-Q4_K_M", "url": "https://huggingface.co/Qwen/Qwen3-VL-2B-Instruct-GGUF/blob/main/Qwen3VL-2B-Instruct-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 1.56, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "ollama qwen3-vl:2b-instruct", "url": "https://ollama.com/library/qwen3-vl:2b-instruct", "precision": "q4_k_m", "format": "ollama", "sizeGB": 1.9, "memoryGB": 8, "runsOn": [ "Ollama" ] } ] }, { "slug": "qwopus3-8-27b-flash-v2", "name": "Qwopus3.8-27B-Flash-V2", "grade": "C", "score": 61, "tasks": [ "agents", "general", "vision" ], "paramsB": 27.8, "params": "27.8B", "context": 262144, "goodFor": "A 27B vision-language fine-tune for long agent loops and coding, trained to reason less per turn, with an MTP head for speculative decoding.", "licence": "Apache-2.0 (metadata only, no LICENSE file); built on Qwen3.8-27B under Apache 2.0", "run": "llama-server -hf Jackrong/Qwopus3.8-27B-Flash-V2-GGUF:Q8_0 --jinja -c 131072", "runBuild": "Qwopus3.8-27B-Flash-V2-MTP-Q8_0.gguf", "url": "/tools/qwopus3-8-27b-flash-v2", "builds": [ { "name": "Qwopus3.8-27B-Flash-V2-MTP-IQ4_XS.gguf", "url": "https://huggingface.co/Jackrong/Qwopus3.8-27B-Flash-V2-GGUF/blob/main/Qwopus3.8-27B-Flash-V2-MTP-IQ4_XS.gguf", "precision": "iq4_xs", "format": "gguf", "sizeGB": 15.42, "memoryGB": 24 }, { "name": "Qwopus3.8-27B-Flash-V2-MTP-Q4_K_S.gguf", "url": "https://huggingface.co/Jackrong/Qwopus3.8-27B-Flash-V2-GGUF/blob/main/Qwopus3.8-27B-Flash-V2-MTP-Q4_K_S.gguf", "precision": "q4_k_s", "format": "gguf", "sizeGB": 15.83, "memoryGB": 24 }, { "name": "Qwopus3.8-27B-Flash-V2-MTP-Q4_K_M.gguf", "url": "https://huggingface.co/Jackrong/Qwopus3.8-27B-Flash-V2-GGUF/blob/main/Qwopus3.8-27B-Flash-V2-MTP-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 16.81, "memoryGB": 24, "runsOn": [ "llama.cpp", "Ollama" ] }, { "name": "Qwopus3.8-27B-Flash-V2-MTP-Q5_K_M.gguf", "url": "https://huggingface.co/Jackrong/Qwopus3.8-27B-Flash-V2-GGUF/blob/main/Qwopus3.8-27B-Flash-V2-MTP-Q5_K_M.gguf", "precision": "q5_k_m", "format": "gguf", "sizeGB": 19.54, "memoryGB": 24 }, { "name": "Qwopus3.8-27B-Flash-V2-MTP-Q6_K.gguf", "url": "https://huggingface.co/Jackrong/Qwopus3.8-27B-Flash-V2-GGUF/blob/main/Qwopus3.8-27B-Flash-V2-MTP-Q6_K.gguf", "precision": "q6_k", "format": "gguf", "sizeGB": 22.43, "memoryGB": 32 }, { "name": "Qwopus3.8-27B-Flash-V2-MTP-Q8_0.gguf", "url": "https://huggingface.co/Jackrong/Qwopus3.8-27B-Flash-V2-GGUF/blob/main/Qwopus3.8-27B-Flash-V2-MTP-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 29.05, "memoryGB": 48, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Qwopus3.8-27B-Flash-V2", "url": "https://huggingface.co/Jackrong/Qwopus3.8-27B-Flash-V2", "precision": "bf16", "format": "safetensors", "sizeGB": 55.58, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)", "Transformers" ] }, { "name": "Qwopus3.8-27B-Flash-V2-MTP-BF16.gguf", "url": "https://huggingface.co/Jackrong/Qwopus3.8-27B-Flash-V2-GGUF/blob/main/Qwopus3.8-27B-Flash-V2-MTP-BF16.gguf", "precision": "bf16", "format": "gguf", "sizeGB": 54.66, "memoryGB": 96 } ] }, { "slug": "recursive-flint", "name": "Recursive Flint", "grade": "C", "score": 61, "tasks": [ "agents", "general", "coding" ], "params": "39.8B total, about 2.5B active", "context": 262144, "goodFor": "Agentic coding on a 32 to 48 GB machine, investigating codebases, fixing bugs and writing tests with multi-turn tool calls and reasoning.", "licence": "Apache-2.0", "run": "llama-server -hf RecursiveSI/flint-int4-gguf --alias flint -ngl all -c 131072 -np 1 -fa on --jinja --temp 1 --top-p 0.95 --top-k 20 --min-p 0 --port 8080", "runBuild": "flint-int4-gguf Q4_1", "url": "/tools/recursive-flint", "builds": [ { "name": "flint-int4", "url": "https://huggingface.co/RecursiveSI/flint-int4", "precision": "int4 (group-64 affine, INT8 embeddings)", "format": "mlx", "sizeGB": 22.6, "memoryGB": 32, "runsOn": [ "rec", "Ollama" ] }, { "name": "flint-int4-gguf Q4_1", "url": "https://huggingface.co/RecursiveSI/flint-int4-gguf", "precision": "q4_1", "format": "gguf", "sizeGB": 26, "memoryGB": 32, "runsOn": [ "llama.cpp" ] }, { "name": "flint-nvfp4-gguf", "url": "https://huggingface.co/RecursiveSI/flint-nvfp4-gguf", "precision": "nvfp4", "format": "gguf", "sizeGB": 22.9, "memoryGB": 32, "runsOn": [ "llama.cpp" ] }, { "name": "ollama flint:latest", "url": "https://ollama.com/library/flint:latest", "precision": "4-bit", "format": "ollama", "sizeGB": 23, "memoryGB": 32, "runsOn": [ "Ollama" ] }, { "name": "lmstudio-community flint-GGUF Q4_K_M", "url": "https://huggingface.co/lmstudio-community/flint-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 24.2, "memoryGB": 32, "runsOn": [ "LM Studio", "llama.cpp" ] }, { "name": "flint", "url": "https://huggingface.co/RecursiveSI/flint", "precision": "bf16", "format": "safetensors", "sizeGB": 79.6, "memoryGB": 96, "runsOn": [ "SGLang 0.5.16 with Recursive's flint_sglang kit" ] } ] }, { "slug": "smollm3-3b", "name": "SmolLM3-3B", "grade": "C", "score": 61, "tasks": [ "agents", "general" ], "paramsB": 3.08, "params": "3.08B", "context": 65536, "goodFor": "Small fully open instruct model with switchable reasoning and tool calling, text only, multilingual.", "licence": "Apache-2.0", "run": "vllm serve HuggingFaceTB/SmolLM3-3B --enable-auto-tool-choice --tool-call-parser=hermes", "runBuild": "SmolLM3-3B", "url": "/tools/smollm3-3b", "builds": [ { "name": "SmolLM3-3B", "url": "https://huggingface.co/HuggingFaceTB/SmolLM3-3B", "precision": "bf16", "format": "safetensors", "sizeGB": 6.2, "memoryGB": 8, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] }, { "name": "SmolLM3-3B-ONNX fp16", "url": "https://huggingface.co/HuggingFaceTB/SmolLM3-3B-ONNX/blob/main/onnx/model_fp16.onnx_data", "precision": "fp16", "format": "onnx", "sizeGB": 6.2, "memoryGB": 8 }, { "name": "SmolLM3-3B-ONNX int8", "url": "https://huggingface.co/HuggingFaceTB/SmolLM3-3B-ONNX/blob/main/onnx/model_int8.onnx_data", "precision": "int8", "format": "onnx", "sizeGB": 3.1, "memoryGB": 8 }, { "name": "SmolLM3-3B-ONNX uint8", "url": "https://huggingface.co/HuggingFaceTB/SmolLM3-3B-ONNX/blob/main/onnx/model_uint8.onnx_data", "precision": "uint8", "format": "onnx", "sizeGB": 3.1, "memoryGB": 8 }, { "name": "SmolLM3-3B-ONNX q4", "url": "https://huggingface.co/HuggingFaceTB/SmolLM3-3B-ONNX/blob/main/onnx/model_q4.onnx_data", "precision": "q4", "format": "onnx", "sizeGB": 2.8, "memoryGB": 8 }, { "name": "SmolLM3-3B-ONNX bnb4", "url": "https://huggingface.co/HuggingFaceTB/SmolLM3-3B-ONNX/blob/main/onnx/model_bnb4.onnx_data", "precision": "bnb4", "format": "onnx", "sizeGB": 2.7, "memoryGB": 8 }, { "name": "SmolLM3-3B-ONNX q4f16", "url": "https://huggingface.co/HuggingFaceTB/SmolLM3-3B-ONNX/blob/main/onnx/model_q4f16.onnx_data", "precision": "q4f16", "format": "onnx", "sizeGB": 2.1, "memoryGB": 8 } ] }, { "slug": "spark-x2-6-4b", "name": "Spark-X2.6-4B", "grade": "C", "score": 61, "tasks": [ "agents", "general" ], "paramsB": 4.11, "params": "4.11B", "context": 1048576, "goodFor": "Small dense text model for coding, tool use and agent loops on a laptop or small GPU, with a 1M-token context and thinking that is off, low or high per request.", "licence": "Apache-2.0", "run": "llama-server -hf XHToken/Spark-X2.6-4B-GGUF:Q8_0 --jinja -c 131072", "runBuild": "Spark-X2.6-4B-Q8_0.gguf", "url": "/tools/spark-x2-6-4b", "builds": [ { "name": "Spark-X2.6-4B-FP8", "url": "https://huggingface.co/XHToken/Spark-X2.6-4B-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 4.47, "memoryGB": 8 }, { "name": "Spark-X2.6-4B-INT8", "url": "https://huggingface.co/XHToken/Spark-X2.6-4B-INT8", "precision": "int8", "format": "safetensors", "sizeGB": 4.47, "memoryGB": 8 }, { "name": "Spark-X2.6-4B-INT4", "url": "https://huggingface.co/XHToken/Spark-X2.6-4B-INT4", "precision": "int4", "format": "safetensors", "sizeGB": 3.43, "memoryGB": 8 }, { "name": "Spark-X2.6-4B-Q4_K_M.gguf", "url": "https://huggingface.co/XHToken/Spark-X2.6-4B-GGUF/blob/main/Spark-X2.6-4B-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 3.19, "memoryGB": 8, "runsOn": [ "llama.cpp", "Ollama" ] }, { "name": "Spark-X2.6-4B-Q8_0.gguf", "url": "https://huggingface.co/XHToken/Spark-X2.6-4B-GGUF/blob/main/Spark-X2.6-4B-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 4.38, "memoryGB": 8, "runsOn": [ "llama.cpp", "Ollama" ] }, { "name": "Spark-X2.6-4B", "url": "https://huggingface.co/XHToken/Spark-X2.6-4B", "precision": "bf16", "format": "safetensors", "sizeGB": 8.24, "memoryGB": 12, "runsOn": [ "SGLang" ] }, { "name": "Spark-X2.6-4B.gguf", "url": "https://huggingface.co/XHToken/Spark-X2.6-4B-GGUF/blob/main/Spark-X2.6-4B.gguf", "precision": "bf16", "format": "gguf", "sizeGB": 8.23, "memoryGB": 12 }, { "name": "ollama SparkLLM/Spark-X2.6-4B:latest", "url": "https://ollama.com/SparkLLM/Spark-X2.6-4B", "precision": "bf16", "format": "ollama", "sizeGB": 8.2, "memoryGB": 12, "runsOn": [ "Ollama" ] } ] }, { "slug": "swift-1-5-qwen3-8-27b", "name": "Swift 1.5 Qwen3.8-27B", "grade": "C", "score": 61, "tasks": [ "agents", "general", "vision" ], "paramsB": 27, "params": "27B", "context": 262144, "goodFor": "A Qwen3.8-27B that thinks in fewer tokens, for coding and agent work with images in, tool calling and adjustable reasoning effort.", "licence": "Swift Open License v1.0 (custom, not an OSI licence) on UkisAI's changes, free for commercial use below US$1M gross annual revenue with a separate Swift Enterprise License above it. Built on Qwen3.8-27B, whose Apache 2.0 still governs the base", "run": "vllm serve ukisai/Swift-1.5-Qwen3.8-27b --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "Swift-1.5-Qwen3.8-27b", "url": "/tools/swift-1-5-qwen3-8-27b", "builds": [ { "name": "Swift-1.5-Qwen3.8-27B-GGUF Q4_K_M", "url": "https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 18.4, "memoryGB": 24, "runsOn": [ "llama.cpp" ] }, { "name": "Swift-1.5-Qwen3.8-27B-GGUF IQ4_XS", "url": "https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GGUF", "precision": "iq4_xs", "format": "gguf", "sizeGB": 16.4, "memoryGB": 24, "runsOn": [ "llama.cpp" ] }, { "name": "Swift-1.5-4bit-MLX", "url": "https://huggingface.co/ukisai/Swift-1.5-4bit-MLX", "precision": "4bit", "format": "mlx", "sizeGB": 15.9, "memoryGB": 24, "runsOn": [ "MLX" ] }, { "name": "Swift-1.5-Qwen3.8-27b-W4A16-AWQ", "url": "https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b-W4A16-AWQ", "precision": "int4", "format": "safetensors", "sizeGB": 19.6, "memoryGB": 24, "runsOn": [ "vLLM" ] }, { "name": "Swift-1.5-Qwen3.8-27b-NVFP4", "url": "https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 21.9, "memoryGB": 32 }, { "name": "Swift-1.5-Qwen3.8-27b-Quark-FP8-dynamic-AMD", "url": "https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b-Quark-FP8-dynamic-AMD", "precision": "fp8", "format": "safetensors", "sizeGB": 31.3, "memoryGB": 48 }, { "name": "Swift-1.5-Qwen3.8-27b", "url": "https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b", "precision": "bf16", "format": "safetensors", "sizeGB": 55.6, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "thinkingcap-qwen3-8-27b", "name": "ThinkingCap (Qwen3.8 27B)", "grade": "C", "score": 61, "tasks": [ "agents", "general", "vision" ], "paramsB": 27, "params": "27B", "context": 262144, "goodFor": "Qwen3.8-27B with shorter reasoning traces, for agent and reasoning work where thinking tokens cost time, with images and video in.", "licence": "PolyForm-Small-Business-1.0.0", "run": "llama-server -hf bottlecapai/ThinkingCap-Qwen3.8-27B-GGUF:Q4_K_M --jinja", "runBuild": "ThinkingCap-Qwen3.8-27B-Q4_K_M", "url": "/tools/thinkingcap-qwen3-8-27b", "builds": [ { "name": "ThinkingCap-Qwen3.8-27B-IQ4_XS", "url": "https://huggingface.co/bottlecapai/ThinkingCap-Qwen3.8-27B-GGUF/blob/main/ThinkingCap-Qwen3.8-27B-IQ4_XS.gguf", "precision": "iq4_xs", "format": "gguf", "sizeGB": 15.5, "memoryGB": 24, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "ThinkingCap-Qwen3.8-27B-Q4_K_M", "url": "https://huggingface.co/bottlecapai/ThinkingCap-Qwen3.8-27B-GGUF/blob/main/ThinkingCap-Qwen3.8-27B-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 17.4, "memoryGB": 24, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "ThinkingCap-Qwen3.8-27B-RCO-4.2bpw", "url": "https://huggingface.co/bottlecapai/ThinkingCap-Qwen3.8-27B-GGUF-RCO-4.2bpw", "precision": "mixed-4.2bpw", "format": "gguf", "sizeGB": 14.3, "memoryGB": 24, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "ThinkingCap-Qwen3.8-27B-NVFP4", "url": "https://huggingface.co/bottlecapai/ThinkingCap-Qwen3.8-27B-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 20.6, "memoryGB": 32 }, { "name": "ThinkingCap-Qwen3.8-27B-NVFP4A4-AWQ", "url": "https://huggingface.co/bottlecapai/ThinkingCap-Qwen3.8-27B-NVFP4A4-AWQ", "precision": "nvfp4", "format": "safetensors", "sizeGB": 23.4, "memoryGB": 32 }, { "name": "ThinkingCap-Qwen3.8-27B-MLX-4bit-DWQ", "url": "https://huggingface.co/bottlecapai/ThinkingCap-Qwen3.8-27B-MLX-4bit-DWQ", "precision": "mlx-4bit", "format": "mlx", "sizeGB": 22.8, "memoryGB": 32, "runsOn": [ "mlx-vlm" ] }, { "name": "ThinkingCap-Qwen3.8-27B-Q6_K", "url": "https://huggingface.co/bottlecapai/ThinkingCap-Qwen3.8-27B-GGUF/blob/main/ThinkingCap-Qwen3.8-27B-Q6_K.gguf", "precision": "q6_k", "format": "gguf", "sizeGB": 23.9, "memoryGB": 32, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "ThinkingCap-Qwen3.8-27B-FP8", "url": "https://huggingface.co/bottlecapai/ThinkingCap-Qwen3.8-27B-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 31.2, "memoryGB": 48 }, { "name": "ThinkingCap-Qwen3.8-27B-Q8_0", "url": "https://huggingface.co/bottlecapai/ThinkingCap-Qwen3.8-27B-GGUF/blob/main/ThinkingCap-Qwen3.8-27B-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 29, "memoryGB": 48, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "ThinkingCap-Qwen3.8-27B", "url": "https://huggingface.co/bottlecapai/ThinkingCap-Qwen3.8-27B", "precision": "bf16", "format": "safetensors", "sizeGB": 55.6, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)", "Transformers" ] }, { "name": "ThinkingCap-Qwen3.8-27B-f16", "url": "https://huggingface.co/bottlecapai/ThinkingCap-Qwen3.8-27B-GGUF/blob/main/ThinkingCap-Qwen3.8-27B-f16.gguf", "precision": "f16", "format": "gguf", "sizeGB": 54.7, "memoryGB": 96, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] } ] }, { "slug": "occamy-1-0", "name": "occamy-1.0", "grade": "C", "score": 61, "tasks": [ "agents", "general", "coding", "vision" ], "activeB": 3, "params": "35B (3B active)", "context": 262144, "goodFor": "Mixture-of-experts agent model for long multi-step work with tools, files, code and office tasks, text and images in, thinking on by default.", "licence": "Apache-2.0", "run": "vllm serve Accio-Lab/occamy-1.0 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "occamy-1.0", "url": "/tools/occamy-1-0", "builds": [ { "name": "occamy-1.0-MLX-mxfp4", "url": "https://huggingface.co/Accio-Lab/occamy-1.0-MLX-mxfp4", "precision": "mxfp4", "format": "mlx", "sizeGB": 18.4, "memoryGB": 24, "runsOn": [ "MLX" ] }, { "name": "occamy-1.0-MLX-4bit", "url": "https://huggingface.co/Accio-Lab/occamy-1.0-MLX-4bit", "precision": "mlx-4bit", "format": "mlx", "sizeGB": 19.5, "memoryGB": 24, "runsOn": [ "MLX" ] }, { "name": "occamy-1.0-IQ4_XS", "url": "https://huggingface.co/Accio-Lab/occamy-1.0-GGUF/blob/main/occamy-1.0-IQ4_XS.gguf", "precision": "iq4_xs", "format": "gguf", "sizeGB": 18.7, "memoryGB": 24, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "occamy-1.0-APEX-Compact", "url": "https://huggingface.co/Accio-Lab/occamy-1.0-APEX-GGUF/blob/main/occamy-1.0-APEX-Compact.gguf", "precision": "mixed-q3-q4", "format": "gguf", "sizeGB": 16.5, "memoryGB": 24, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "occamy-1.0-NVFP4", "url": "https://huggingface.co/Accio-Lab/occamy-1.0-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 23.9, "memoryGB": 32 }, { "name": "occamy-1.0-Q4_K_M", "url": "https://huggingface.co/Accio-Lab/occamy-1.0-GGUF/blob/main/occamy-1.0-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 21.2, "memoryGB": 32, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "occamy-1.0-FP8", "url": "https://huggingface.co/Accio-Lab/occamy-1.0-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 36.6, "memoryGB": 48, "runsOn": [ "vLLM", "SGLang" ] }, { "name": "occamy-1.0-Q8_0", "url": "https://huggingface.co/Accio-Lab/occamy-1.0-GGUF/blob/main/occamy-1.0-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 36.9, "memoryGB": 48, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "occamy-1.0", "url": "https://huggingface.co/Accio-Lab/occamy-1.0", "precision": "bf16", "format": "safetensors", "sizeGB": 70.2, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)", "Transformers" ] } ] }, { "slug": "nanbeige-4-3-33b", "name": "Nanbeige4.3-33B", "grade": "C", "score": 60.8, "tasks": [ "agents", "general", "coding" ], "paramsB": 33, "params": "33B", "context": 524288, "goodFor": "A cousin for long agentic coding and office tasks on a 32 GB GPU at INT4 or about 96 GB at BF16, with a 512K context and per-request thinking.", "licence": "Apache-2.0 (Hub metadata only; the repository has no LICENSE file)", "run": "vllm serve Nanbeige/Nanbeige4.3-33B-INT4 --enable-auto-tool-choice --reasoning-parser qwen3 --tool-call-parser qwen3_coder", "runBuild": "Nanbeige4.3-33B-INT4", "url": "/tools/nanbeige-4-3-33b", "builds": [ { "name": "Nanbeige4.3-33B-INT4", "url": "https://huggingface.co/Nanbeige/Nanbeige4.3-33B-INT4", "precision": "int4", "format": "safetensors", "sizeGB": 22.57, "memoryGB": 32, "runsOn": [ "vLLM", "SGLang" ] }, { "name": "Nanbeige4.3-33B-FP8", "url": "https://huggingface.co/Nanbeige/Nanbeige4.3-33B-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 35.54, "memoryGB": 48, "runsOn": [ "vLLM", "SGLang" ] }, { "name": "Nanbeige4.3-33B (BF16)", "url": "https://huggingface.co/Nanbeige/Nanbeige4.3-33B", "precision": "bf16", "format": "safetensors", "sizeGB": 66.26, "memoryGB": 96, "runsOn": [ "vLLM", "SGLang", "Transformers" ] } ] }, { "slug": "qwen3-5-9b", "name": "Qwen3.5-9B", "grade": "C", "score": 60.5, "tasks": [ "agents", "general", "vision" ], "paramsB": 9, "params": "9B", "context": 262144, "goodFor": "Dense vision-language model for local agents: text, images and video in, tool calling, thinking on by default.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3.5-9B --port 8000 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "Qwen3.5-9B", "url": "/tools/qwen3-5-9b", "builds": [ { "name": "ollama qwen3.5:9b", "url": "https://ollama.com/library/qwen3.5:9b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 6.6, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "Qwen3.5-9B", "url": "https://huggingface.co/Qwen/Qwen3.5-9B", "precision": "bf16", "format": "safetensors", "sizeGB": 19.3, "memoryGB": 24, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "gemma-4-e2b-it", "name": "Gemma 4 E2B IT", "grade": "C", "score": 60.4, "tasks": [ "agents", "general", "vision" ], "paramsB": 5.1, "params": "5.1B", "context": 131072, "goodFor": "Small on-device assistant and agent model, 2.3B effective, with tool calling and optional thinking, taking text, images, video frames and short audio clips for transcription and translation.", "licence": "Apache-2.0", "run": "llama-server -hf google/gemma-4-E2B-it-qat-q4_0-gguf --hf-file gemma-4-E2B_q4_0-it.gguf --jinja -c 131072", "runBuild": "gemma-4-E2B_q4_0-it", "url": "/tools/gemma-4-e2b-it", "builds": [ { "name": "gemma-4-E2B_q4_0-it", "url": "https://huggingface.co/google/gemma-4-E2B-it-qat-q4_0-gguf/blob/main/gemma-4-E2B_q4_0-it.gguf", "precision": "q4_0", "format": "gguf", "sizeGB": 4.34, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)", "LM Studio" ] }, { "name": "gemma-4-E2B-it-qat-mobile-ct", "url": "https://huggingface.co/google/gemma-4-E2B-it-qat-mobile-ct", "precision": "wNa8o8", "format": "safetensors", "sizeGB": 2.7, "memoryGB": 8 }, { "name": "gemma-4-E2B-it-qat-mobile-transformers", "url": "https://huggingface.co/google/gemma-4-E2B-it-qat-mobile-transformers", "precision": "wNa8o8", "format": "safetensors", "sizeGB": 2.49, "memoryGB": 8 }, { "name": "ollama gemma4:e2b", "url": "https://ollama.com/library/gemma4:e2b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 4.6, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "lmstudio google/gemma-4-e2b", "url": "https://lmstudio.ai/models/google/gemma-4-e2b", "precision": "q4_k_m", "format": "gguf", "sizeGB": 4.2, "memoryGB": 8, "runsOn": [ "LM Studio" ] }, { "name": "gemma-4-E2B-it-qat-w4a16-ct", "url": "https://huggingface.co/google/gemma-4-E2B-it-qat-w4a16-ct", "precision": "w4a16", "format": "safetensors", "sizeGB": 8.35, "memoryGB": 12 }, { "name": "gemma-4-E2B-it", "url": "https://huggingface.co/google/gemma-4-E2B-it", "precision": "bf16", "format": "safetensors", "sizeGB": 10.28, "memoryGB": 16, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "cohere-transcribe-03-2026", "name": "Cohere Transcribe 03-2026", "grade": "C", "score": 60.3, "tasks": [ "stt" ], "paramsB": 2.07, "params": "2.07B", "goodFor": "Accurate offline and batch transcription in 14 languages when the language is known, on a small GPU.", "licence": "Apache-2.0", "run": "vllm serve CohereLabs/cohere-transcribe-03-2026 --trust-remote-code", "runBuild": "cohere-transcribe-03-2026", "url": "/tools/cohere-transcribe-03-2026", "builds": [ { "name": "cohere-transcribe-03-2026", "url": "https://huggingface.co/CohereLabs/cohere-transcribe-03-2026", "precision": "bf16", "format": "safetensors", "sizeGB": 4.13, "memoryGB": 8, "runsOn": [ "vLLM", "Transformers" ] } ] }, { "slug": "glm-4-5-air", "name": "GLM-4.5-Air", "grade": "C", "score": 60.3, "tasks": [ "agents", "general" ], "paramsB": 106, "activeB": 12, "params": "106B", "context": 131072, "goodFor": "Mid-size mixture-of-experts reasoning model for English and Chinese agent, coding and tool-calling work, with thinking that turns off per request.", "licence": "MIT", "run": "python3 -m sglang.launch_server --model-path zai-org/GLM-4.5-Air-FP8 --tp-size 4 --tool-call-parser glm45 --reasoning-parser glm45 --served-model-name glm-4.5-air --port 8000", "runBuild": "GLM-4.5-Air-FP8", "url": "/tools/glm-4-5-air", "builds": [ { "name": "GLM-4.5-Air-FP8", "url": "https://huggingface.co/zai-org/GLM-4.5-Air-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 112.6, "memoryGB": 192 }, { "name": "GLM-4.5-Air", "url": "https://huggingface.co/zai-org/GLM-4.5-Air", "precision": "bf16", "format": "safetensors", "sizeGB": 221, "memoryGB": 512 } ] }, { "slug": "nemotron-3-5-asr-streaming-0-6b", "name": "Nemotron 3.5 ASR Streaming 0.6B", "grade": "C", "score": 60.3, "tasks": [ "stt" ], "paramsB": 0.64, "params": "0.64B", "goodFor": "Low-latency streaming transcription in 32 languages from one small model, with automatic language detection, for voice agents and many parallel streams.", "licence": "OpenMDW-1.1", "run": "hf download nvidia/nemotron-3.5-asr-streaming-0.6b nemotron-3.5-asr-streaming-0.6b.q8_0.gguf --local-dir models \u0026\u0026 nemo-speech transcribe audio.wav --model models/nemotron-3.5-asr-streaming-0.6b.q8_0.gguf --language auto", "runBuild": "nemotron-3.5-asr-streaming-0.6b.q8_0.gguf", "url": "/tools/nemotron-3-5-asr-streaming-0-6b", "builds": [ { "name": "nemotron-3.5-asr-streaming-0.6b (safetensors)", "url": "https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b/blob/main/model.safetensors", "precision": "fp32", "format": "safetensors", "sizeGB": 2.55, "memoryGB": 8, "runsOn": [ "Transformers" ] }, { "name": "nemotron-3.5-asr-streaming-0.6b.nemo", "url": "https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b/blob/main/nemotron-3.5-asr-streaming-0.6b.nemo", "precision": "fp32", "format": "nemo", "sizeGB": 2.37, "memoryGB": 8 }, { "name": "nemotron-3.5-asr-streaming-0.6b.q8_0.gguf", "url": "https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b/blob/main/nemotron-3.5-asr-streaming-0.6b.q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 0.74, "memoryGB": 8 } ] }, { "slug": "granite-4-2-3b", "name": "Granite 4.2 3B", "grade": "C", "score": 59.8, "tasks": [ "agents", "general" ], "paramsB": 3.66, "params": "3.66B", "context": 131072, "goodFor": "Small dense reasoning model for tool calling, coding and multilingual chat, with thinking that is on, off or short per request.", "licence": "Apache-2.0", "run": "ollama run granite4.2:3b", "runBuild": "ollama granite4.2:3b", "url": "/tools/granite-4-2-3b", "builds": [ { "name": "granite-4.2-3b-fp8", "url": "https://huggingface.co/ibm-granite/granite-4.2-3b-fp8", "precision": "fp8", "format": "safetensors", "sizeGB": 4.19, "memoryGB": 8, "runsOn": [ "vLLM" ] }, { "name": "granite-4.2-3b-nvfp4", "url": "https://huggingface.co/ibm-granite/granite-4.2-3b-nvfp4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 2.81, "memoryGB": 8, "runsOn": [ "vLLM" ] }, { "name": "granite-4.2-3b-mxfp4", "url": "https://huggingface.co/ibm-granite/granite-4.2-3b-mxfp4", "precision": "mxfp4", "format": "safetensors", "sizeGB": 2.71, "memoryGB": 8, "runsOn": [ "vLLM" ] }, { "name": "granite-4.2-3b-Q8_0", "url": "https://huggingface.co/ibm-granite/granite-4.2-3b-GGUF/blob/main/granite-4.2-3b-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 3.89, "memoryGB": 8, "runsOn": [ "llama.cpp" ] }, { "name": "granite-4.2-3b-Q4_K_M", "url": "https://huggingface.co/ibm-granite/granite-4.2-3b-GGUF/blob/main/granite-4.2-3b-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 2.24, "memoryGB": 8, "runsOn": [ "llama.cpp" ] }, { "name": "granite-4.2-3b-q4-mlx", "url": "https://huggingface.co/ibm-granite/granite-4.2-3b-q4-mlx", "precision": "q4", "format": "mlx", "sizeGB": 2.07, "memoryGB": 8, "runsOn": [ "mlx-lm" ] }, { "name": "ollama granite4.2:3b", "url": "https://ollama.com/library/granite4.2:3b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 2.2, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "granite-4.2-3b", "url": "https://huggingface.co/ibm-granite/granite-4.2-3b", "precision": "bf16", "format": "safetensors", "sizeGB": 7.33, "memoryGB": 12, "runsOn": [ "vLLM", "SGLang", "Transformers" ] } ] }, { "slug": "minicpm5-2b", "name": "MiniCPM5-2B", "grade": "C", "score": 59.7, "tasks": [ "agents", "general" ], "paramsB": 2.5, "params": "2.5B", "context": 131072, "goodFor": "Small on-device language model with thinking and XML tool calls, text only.", "licence": "Apache-2.0", "run": "ollama run openbmb/minicpm5-2b", "runBuild": "ollama openbmb/minicpm5-2b:latest", "url": "/tools/minicpm5-2b", "builds": [ { "name": "MiniCPM5-2B", "url": "https://huggingface.co/openbmb/MiniCPM5-2B", "precision": "bf16", "format": "safetensors", "sizeGB": 5, "memoryGB": 8, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] }, { "name": "MiniCPM5-2B-F16", "url": "https://huggingface.co/openbmb/MiniCPM5-2B-GGUF/blob/main/MiniCPM5-2B-F16.gguf", "precision": "f16", "format": "gguf", "sizeGB": 5, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)", "LM Studio" ] }, { "name": "MiniCPM5-2B-Q8_0", "url": "https://huggingface.co/openbmb/MiniCPM5-2B-GGUF/blob/main/MiniCPM5-2B-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 2.7, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)", "LM Studio" ] }, { "name": "MiniCPM5-2B-Q4_K_M", "url": "https://huggingface.co/openbmb/MiniCPM5-2B-GGUF/blob/main/MiniCPM5-2B-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 1.6, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)", "LM Studio" ] }, { "name": "MiniCPM5-2B-MLX", "url": "https://huggingface.co/openbmb/MiniCPM5-2B-MLX", "precision": "4bit", "format": "mlx", "sizeGB": 1.4, "memoryGB": 8 }, { "name": "MiniCPM5-2B-GPTQ", "url": "https://huggingface.co/openbmb/MiniCPM5-2B-GPTQ", "precision": "gptq-int4", "format": "safetensors", "sizeGB": 2.1, "memoryGB": 8 }, { "name": "ollama openbmb/minicpm5-2b:latest", "url": "https://ollama.com/openbmb/minicpm5-2b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 1.6, "memoryGB": 8, "runsOn": [ "Ollama" ] } ] }, { "slug": "cosmos-reason2-2b", "name": "Cosmos-Reason2-2B", "grade": "C", "score": 59.5, "tasks": [ "general", "vision" ], "paramsB": 2.44, "params": "2.44B", "context": 262144, "goodFor": "Small vision language model for physical AI, reasoning over video and images for robots, driving and smart spaces, with point and box outputs.", "licence": "NVIDIA Open Model License (custom, commercial use allowed, not an OSI licence); built on Qwen3-VL-2B-Instruct under Apache 2.0", "run": "vllm serve nvidia/Cosmos-Reason2-2B --allowed-local-media-path \"$(pwd)/media\" --max-model-len 16384 --media-io-kwargs '{\"video\": {\"num_frames\": -1}}' --reasoning-parser qwen3 --port 8000", "runBuild": "Cosmos-Reason2-2B", "url": "/tools/cosmos-reason2-2b", "builds": [ { "name": "Cosmos-Reason2-2B", "url": "https://huggingface.co/nvidia/Cosmos-Reason2-2B", "precision": "bf16", "format": "safetensors", "sizeGB": 4.9, "memoryGB": 8 } ] }, { "slug": "jev-27b-vl", "name": "JEV-27B-VL", "grade": "C", "score": 59.5, "tasks": [ "decide" ], "paramsB": 27, "params": "27B", "context": 262144, "goodFor": "Typed decisions with calibrated probabilities over text and images, yes or no, one of up to 256 options or a 0 to 5 rating, in one forward pass, with optional thinking from the base model for hard cases.", "licence": "Apache-2.0 for AutoTrust's System 1 adapter and decision head. The base Qwen3.8-27B weights in the repository are Alibaba's, unchanged, also Apache-2.0", "run": "vllm serve JEV-27B-VL --served-model-name autotrust/JEV-27B-VL --enable-lora --max-lora-rank 32 --lora-modules jev-decision=JEV-27B-VL/adapter_vllm --logprobs-mode processed_logprobs --max-model-len 32768 --enable-prefix-caching --mamba-cache-mode align --limit-mm-per-prompt '{\"image\": 8}' --max-num-seqs 8 --trust-request-chat-template", "runBuild": "JEV-27B-VL", "url": "/tools/jev-27b-vl", "builds": [ { "name": "JEV-27B-VL", "url": "https://huggingface.co/autotrust/JEV-27B-VL", "precision": "bf16", "format": "safetensors", "sizeGB": 56, "memoryGB": 96, "runsOn": [ "vLLM" ] } ] }, { "slug": "ming-3-0-tts-0-5b", "name": "Ming-3.0-tts-0.5B", "grade": "C", "score": 59.3, "tasks": [ "tts" ], "paramsB": 0.5, "params": "0.5B", "goodFor": "Text to speech with voice cloning, voice design, built-in voices and inline paralinguistic tags in seven languages.", "licence": "MIT", "run": "git clone https://github.com/inclusionAI/Ming-3.0-tts.git \u0026\u0026 cd Ming-3.0-tts \u0026\u0026 pip install -r requirements.txt \u0026\u0026 python infer.py --model_path inclusionAI/Ming-3.0-tts-0.5B --task IP --text \"The garden will remain open until sunset.\" --ip US_James --use_zero_spk_emb --no_text_normalization --output output/tts.wav", "runBuild": "Ming-3.0-tts-0.5B", "url": "/tools/ming-3-0-tts-0-5b", "builds": [ { "name": "Ming-3.0-tts-0.5B", "url": "https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B", "precision": "bf16", "format": "safetensors", "sizeGB": 3, "memoryGB": 8 } ] }, { "slug": "qwen3-tts-1-7b-customvoice", "name": "Qwen3-TTS 12Hz 1.7B CustomVoice", "grade": "C", "score": 59.2, "tasks": [ "tts" ], "params": "1.92B (main checkpoint) plus the 12 Hz speech tokenizer", "context": 32768, "goodFor": "Local text-to-speech with nine preset voices whose emotion and style follow an instruction, in ten languages and two Chinese dialects.", "licence": "Apache-2.0", "run": "pip install -U qwen-tts \u0026\u0026 qwen-tts-demo Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice --ip 127.0.0.1 --port 8000", "runBuild": "Qwen3-TTS-12Hz-1.7B-CustomVoice", "url": "/tools/qwen3-tts-1-7b-customvoice", "builds": [ { "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice", "url": "https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice", "precision": "bf16", "format": "safetensors", "sizeGB": 4.52, "memoryGB": 8 } ] }, { "slug": "minimax-m2-5", "name": "MiniMax-M2.5", "grade": "C", "score": 59, "tasks": [ "agents", "general", "coding" ], "paramsB": 228.7, "params": "228.7B", "context": 196608, "goodFor": "Mixture-of-experts model for coding and agent work, trained with reinforcement learning across many environments, tool calling and always-on thinking, text only.", "licence": "MiniMax Model License (model) and modified MIT (code)", "run": "SAFETENSORS_FAST_GPU=1 vllm serve MiniMaxAI/MiniMax-M2.5 --trust-remote-code --tensor-parallel-size 4 --enable-auto-tool-choice --tool-call-parser minimax_m2 --reasoning-parser minimax_m2_append_think", "runBuild": "MiniMax-M2.5", "url": "/tools/minimax-m2-5", "builds": [ { "name": "MiniMax-M2.5", "url": "https://huggingface.co/MiniMaxAI/MiniMax-M2.5", "precision": "fp8", "format": "safetensors", "sizeGB": 230.1, "memoryGB": 512 } ] }, { "slug": "ornith-1-5-397b", "name": "Ornith-1.5-397B", "grade": "C", "score": 59, "tasks": [ "agents", "general", "coding", "vision" ], "paramsB": 397, "params": "397B", "context": 262144, "goodFor": "Flagship coding and terminal-agent mixture-of-experts model with tool calling and thinking on by default, text and images in, for a multi-GPU server.", "licence": "MIT", "run": "ollama run ornith-1.5:397b", "runBuild": "ollama ornith-1.5:397b", "url": "/tools/ornith-1-5-397b", "builds": [ { "name": "Ornith-1.5-397B-FP8", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-397B-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 418.4, "memoryGB": 512 }, { "name": "Ornith-1.5-397B-NVFP4", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-397B-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 319.5, "memoryGB": 512 }, { "name": "Ornith-1.5-397B-Q6_K", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-397B-GGUF/blob/main/Ornith-1.5-397B-Q6_K.gguf", "precision": "q6_k", "format": "gguf", "sizeGB": 331, "memoryGB": 512 }, { "name": "Ornith-1.5-397B-Q5_K_M", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-397B-GGUF/blob/main/Ornith-1.5-397B-Q5_K_M.gguf", "precision": "q5_k_m", "format": "gguf", "sizeGB": 286.3, "memoryGB": 512 }, { "name": "Ornith-1.5-397B-Q4_K_M", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-397B-GGUF/blob/main/Ornith-1.5-397B-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 244.3, "memoryGB": 512 }, { "name": "ollama ornith-1.5:397b", "url": "https://ollama.com/library/ornith-1.5:397b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 242, "memoryGB": 512, "runsOn": [ "Ollama" ] }, { "name": "Ornith-1.5-397B", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-397B", "precision": "bf16", "format": "safetensors", "sizeGB": 806.8, "memoryGB": 1024 }, { "name": "Ornith-1.5-397B-Q8_0", "url": "https://huggingface.co/ornith-ai/Ornith-1.5-397B-GGUF/blob/main/Ornith-1.5-397B-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 428.5, "memoryGB": 1024 } ] }, { "slug": "xiaomi-ocr-0", "name": "Xiaomi-OCR-0", "grade": "C", "score": 59, "tasks": [ "docs" ], "params": "0.87B (873,438,784 by the Hub's count; 0.8B on the card)", "context": 16384, "goodFor": "Reading document images on small hardware, scanned, photographed or skewed pages to Markdown with LaTeX formulas and tables, receipts and forms to JSON fields, and short answers about a page, locally and under a permissive licence.", "licence": "Apache-2.0 (card metadata and card text; the Hugging Face repository has no LICENSE file, and the GitHub code repository's LICENSE is the Apache 2.0 text, copyright 2026 Xiaomi Corporation)", "run": "vllm serve xiaomi-research/Xiaomi-OCR-0 --host 127.0.0.1 --port 8000 --max-model-len 16384", "runBuild": "Xiaomi-OCR-0 (BF16 safetensors)", "url": "/tools/xiaomi-ocr-0", "builds": [ { "name": "Xiaomi-OCR-0 (BF16 safetensors)", "url": "https://huggingface.co/xiaomi-research/Xiaomi-OCR-0/blob/main/model.safetensors", "precision": "bf16", "format": "safetensors", "sizeGB": 1.75, "memoryGB": 8, "runsOn": [ "Transformers", "vLLM", "SGLang" ] } ] }, { "slug": "deepseek-v3-2", "name": "DeepSeek-V3.2", "grade": "C", "score": 58.6, "tasks": [ "agents", "general" ], "paramsB": 685, "params": "685B", "context": 131072, "goodFor": "Large text reasoning model trained for tool use and search agents, with thinking and chat modes chosen per request.", "licence": "MIT", "run": "vllm serve deepseek-ai/DeepSeek-V3.2 --tensor-parallel-size 8 --tokenizer-mode deepseek_v32 --enable-auto-tool-choice --tool-call-parser deepseek_v32 --reasoning-parser deepseek_v3", "runBuild": "DeepSeek-V3.2", "url": "/tools/deepseek-v3-2", "builds": [ { "name": "DeepSeek-V3.2", "url": "https://huggingface.co/deepseek-ai/DeepSeek-V3.2", "precision": "fp8", "format": "safetensors", "sizeGB": 689.5, "memoryGB": 1024 } ] }, { "slug": "vieneu-tts-v3-turbo", "name": "VieNeu-TTS v3 Turbo", "grade": "C", "score": 58.5, "tasks": [ "tts" ], "paramsB": 0.13, "params": "0.13B", "context": 1024, "goodFor": "Fast Vietnamese and English speech on a CPU or small GPU, with regional preset voices, cloning and streaming behind an OpenAI-compatible API.", "licence": "Apache-2.0", "run": "pip install vieneu \u0026\u0026 python -c \"from vieneu import Vieneu; t = Vieneu(); t.save(t.infer('Xin chào!', voice='Mai Anh'), 'out.wav')\"", "runBuild": "VieNeu-TTS-v3-Turbo onnx_update/ (FP32)", "url": "/tools/vieneu-tts-v3-turbo", "builds": [ { "name": "VieNeu-TTS-v3-Turbo update/ (safetensors)", "url": "https://huggingface.co/pnnbao-ump/VieNeu-TTS-v3-Turbo/tree/main/update", "precision": "bf16", "format": "safetensors", "sizeGB": 0.25, "memoryGB": 8 }, { "name": "VieNeu-TTS-v3-Turbo onnx_update/ (FP32)", "url": "https://huggingface.co/pnnbao-ump/VieNeu-TTS-v3-Turbo/tree/main/onnx_update", "precision": "fp32", "format": "onnx", "sizeGB": 0.48, "memoryGB": 8 }, { "name": "VieNeu-TTS-v3-Turbo onnx_int8/ (INT8)", "url": "https://huggingface.co/pnnbao-ump/VieNeu-TTS-v3-Turbo/tree/main/onnx_int8", "precision": "int8", "format": "onnx", "sizeGB": 0.17, "memoryGB": 8 }, { "name": "vieneu-v3-turbo-q8_0.gguf", "url": "https://huggingface.co/pnnbao-ump/VieNeu-TTS-v3-Turbo/blob/main/gguf/vieneu-v3-turbo-q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 0.19, "memoryGB": 8 } ] }, { "slug": "qwen3-vl-30b-a3b-instruct", "name": "Qwen3-VL-30B-A3B-Instruct", "grade": "C", "score": 58, "tasks": [ "agents", "general", "vision" ], "params": "31.1B (3B active)", "context": 262144, "goodFor": "Vision-language mixture of experts for images, documents, screens and video with text out, about 3B of 31B parameters active per token, no thinking mode in this Instruct edition.", "licence": "Apache-2.0", "run": "llama-server -hf Qwen/Qwen3-VL-30B-A3B-Instruct-GGUF:Q4_K_M --jinja -c 32768", "runBuild": "Qwen3VL-30B-A3B-Instruct-Q4_K_M", "url": "/tools/qwen3-vl-30b-a3b-instruct", "builds": [ { "name": "Qwen3VL-30B-A3B-Instruct-Q4_K_M", "url": "https://huggingface.co/Qwen/Qwen3-VL-30B-A3B-Instruct-GGUF/blob/main/Qwen3VL-30B-A3B-Instruct-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 18.6, "memoryGB": 24, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "ollama qwen3-vl:30b-a3b-instruct", "url": "https://ollama.com/library/qwen3-vl:30b-a3b-instruct", "precision": "q4_k_m", "format": "ollama", "sizeGB": 20, "memoryGB": 24, "runsOn": [ "Ollama" ] }, { "name": "Qwen3-VL-30B-A3B-Instruct-FP8", "url": "https://huggingface.co/Qwen/Qwen3-VL-30B-A3B-Instruct-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 32.3, "memoryGB": 48 }, { "name": "Qwen3VL-30B-A3B-Instruct-Q8_0", "url": "https://huggingface.co/Qwen/Qwen3-VL-30B-A3B-Instruct-GGUF/blob/main/Qwen3VL-30B-A3B-Instruct-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 32.5, "memoryGB": 48, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Qwen3-VL-30B-A3B-Instruct", "url": "https://huggingface.co/Qwen/Qwen3-VL-30B-A3B-Instruct", "precision": "bf16", "format": "safetensors", "sizeGB": 62.2, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] }, { "name": "Qwen3VL-30B-A3B-Instruct-F16", "url": "https://huggingface.co/Qwen/Qwen3-VL-30B-A3B-Instruct-GGUF/blob/main/Qwen3VL-30B-A3B-Instruct-F16-split-00001-of-00002.gguf", "precision": "f16", "format": "gguf", "sizeGB": 61.1, "memoryGB": 96, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] } ] }, { "slug": "qwen3-vl-embedding-2b", "name": "Qwen3-VL-Embedding-2B", "grade": "C", "score": 57.7, "tasks": [ "embed" ], "paramsB": 2.13, "params": "2.13B", "context": 32768, "goodFor": "Multimodal retrieval and clustering, one vector for text, images, screenshots, video or a mix of them, in 30 or more languages.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3-VL-Embedding-2B --runner pooling --max-model-len 8192", "runBuild": "Qwen3-VL-Embedding-2B", "url": "/tools/qwen3-vl-embedding-2b", "builds": [ { "name": "Qwen3-VL-Embedding-2B", "url": "https://huggingface.co/Qwen/Qwen3-VL-Embedding-2B", "precision": "bf16", "format": "safetensors", "sizeGB": 4.27, "memoryGB": 8, "runsOn": [ "Sentence Transformers", "vLLM", "SGLang" ] } ] }, { "slug": "minimax-m2-7", "name": "MiniMax-M2.7", "grade": "C", "score": 57.6, "tasks": [ "agents", "general", "coding" ], "paramsB": 228.7, "params": "228.7B", "context": 204800, "goodFor": "Mixture-of-experts model for coding, agent harnesses and office documents, with parallel tool calls and always-on thinking, text only, for non-commercial use unless MiniMax authorises otherwise.", "licence": "MiniMax M2.7 non-commercial licence (MIT-style terms, commercial use needs MiniMax's written authorisation)", "run": "SAFETENSORS_FAST_GPU=1 vllm serve MiniMaxAI/MiniMax-M2.7 --trust-remote-code --tensor-parallel-size 4 --enable-auto-tool-choice --tool-call-parser minimax_m2 --reasoning-parser minimax_m2_append_think", "runBuild": "MiniMax-M2.7", "url": "/tools/minimax-m2-7", "builds": [ { "name": "LM Studio minimax/minimax-m2.7 (Q4_K_M GGUF)", "url": "https://huggingface.co/lmstudio-community/MiniMax-M2.7-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 138.3, "memoryGB": 192, "runsOn": [ "LM Studio", "llama.cpp" ] }, { "name": "MiniMax-M2.7", "url": "https://huggingface.co/MiniMaxAI/MiniMax-M2.7", "precision": "fp8", "format": "safetensors", "sizeGB": 230.2, "memoryGB": 512 } ] }, { "slug": "qwen3-coder-next", "name": "Qwen3-Coder-Next", "grade": "C", "score": 57.6, "tasks": [ "agents", "general", "coding" ], "params": "80B total, 3B active", "context": 262144, "goodFor": "Fast coding-agent model for local development, text only, trained for tool calling across many agent harness formats, no thinking mode.", "licence": "Apache-2.0", "run": "hf download Qwen/Qwen3-Coder-Next-GGUF --include \"Qwen3-Coder-Next-Q4_K_M/*\" --local-dir . \u0026\u0026 llama-server -m Qwen3-Coder-Next-Q4_K_M/Qwen3-Coder-Next-Q4_K_M-00001-of-00004.gguf --jinja -ngl 99 -fa on -c 131072 --temp 1.0 --top-k 40 --top-p 0.95 --min-p 0", "runBuild": "Qwen3-Coder-Next-GGUF Q4_K_M", "url": "/tools/qwen3-coder-next", "builds": [ { "name": "Qwen3-Coder-Next-GGUF Q4_K_M", "url": "https://huggingface.co/Qwen/Qwen3-Coder-Next-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 48.4, "memoryGB": 64, "runsOn": [ "llama.cpp" ] }, { "name": "ollama qwen3-coder-next:latest", "url": "https://ollama.com/library/qwen3-coder-next:latest", "precision": "q4_k_m", "format": "ollama", "sizeGB": 52, "memoryGB": 64, "runsOn": [ "Ollama" ] }, { "name": "lmstudio-community Qwen3-Coder-Next-MLX-4bit", "url": "https://huggingface.co/lmstudio-community/Qwen3-Coder-Next-MLX-4bit", "precision": "mlx-4bit", "format": "mlx", "sizeGB": 44.9, "memoryGB": 64, "runsOn": [ "LM Studio (Apple silicon)" ] }, { "name": "Qwen3-Coder-Next-FP8", "url": "https://huggingface.co/Qwen/Qwen3-Coder-Next-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 80.4, "memoryGB": 128 }, { "name": "Qwen3-Coder-Next-GGUF Q8_0", "url": "https://huggingface.co/Qwen/Qwen3-Coder-Next-GGUF", "precision": "q8_0", "format": "gguf", "sizeGB": 84.8, "memoryGB": 128, "runsOn": [ "llama.cpp" ] }, { "name": "Qwen3-Coder-Next", "url": "https://huggingface.co/Qwen/Qwen3-Coder-Next", "precision": "bf16", "format": "safetensors", "sizeGB": 159.4, "memoryGB": 192, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "joey-1", "name": "joey-1", "grade": "C", "score": 57.1, "tasks": [ "decide" ], "paramsB": 2, "params": "2B", "context": 16384, "goodFor": "Cheap typed decisions with calibrated probabilities over text and images, several choice, yes or no and score questions about one state in one prefill, for routing, triage and guardrails on a small GPU.", "licence": "MIT, named in the card's metadata (license: mit) since the upload of 27 October 2026, with no LICENSE file. The base Qwen/Qwen3.5-2B is Apache-2.0", "run": "bash setup.sh \u0026\u0026 python -c 'from qev.engine import DecisionEngine; DecisionEngine(\"baselabs/joey-1\", device=\"cuda:0\")'", "runBuild": "joey-1", "url": "/tools/joey-1", "builds": [ { "name": "joey-1", "url": "https://huggingface.co/baselabs/joey-1", "precision": "bf16", "format": "safetensors", "sizeGB": 5.5, "memoryGB": 8, "runsOn": [ "OneJev qev with serve.patch" ] } ] }, { "slug": "omnivoice", "name": "OmniVoice", "grade": "C", "score": 56.9, "tasks": [ "tts" ], "params": "0.61B (main checkpoint) plus the Higgs Audio v2 tokenizer", "goodFor": "Zero-shot voice cloning and attribute-designed voices in over 600 languages, for non-commercial use.", "licence": "CC-BY-NC (weights, per the card); Apache-2.0 (code)", "run": "pip install omnivoice \u0026\u0026 omnivoice-infer --model k2-fsa/OmniVoice --text \"This is a test for text to speech.\" --instruct \"female, low pitch, british accent\" --output hello.wav", "runBuild": "OmniVoice", "url": "/tools/omnivoice", "builds": [ { "name": "OmniVoice", "url": "https://huggingface.co/k2-fsa/OmniVoice", "precision": "fp32", "format": "safetensors", "sizeGB": 3.27, "memoryGB": 8 } ] }, { "slug": "nemotron-3-nano-4b", "name": "Nemotron 3 Nano 4B", "grade": "C", "score": 56.8, "tasks": [ "agents", "general" ], "paramsB": 4, "params": "4B", "context": 262144, "goodFor": "Small English reasoning and tool-calling model for agents on edge devices and laptops, with a thinking switch and 262,144 tokens of context.", "licence": "NVIDIA Nemotron Open Model License", "run": "llama-server -hf nvidia/NVIDIA-Nemotron-3-Nano-4B-GGUF --jinja -c 131072", "runBuild": "NVIDIA-Nemotron-3-Nano-4B-GGUF (Q4_K_M)", "url": "/tools/nemotron-3-nano-4b", "builds": [ { "name": "NVIDIA-Nemotron-3-Nano-4B-FP8", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 5.28, "memoryGB": 8 }, { "name": "NVIDIA-Nemotron-3-Nano-4B-GGUF (Q4_K_M)", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-4B-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 2.84, "memoryGB": 8, "runsOn": [ "llama.cpp" ] }, { "name": "ollama nemotron-3-nano:4b", "url": "https://ollama.com/library/nemotron-3-nano:4b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 2.8, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "lmstudio-community NVIDIA-Nemotron-3-Nano-4B-GGUF (Q4_K_M)", "url": "https://huggingface.co/lmstudio-community/NVIDIA-Nemotron-3-Nano-4B-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 2.84, "memoryGB": 8, "runsOn": [ "LM Studio" ] }, { "name": "NVIDIA-Nemotron-3-Nano-4B-BF16", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16", "precision": "bf16", "format": "safetensors", "sizeGB": 7.96, "memoryGB": 12, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "laguna-m-1", "name": "Laguna M.1", "grade": "C", "score": 56.7, "tasks": [ "agents", "general", "coding" ], "params": "225.8B total, 23B active", "context": 262144, "goodFor": "Agentic coding on a multi-GPU server, with tool calling and reasoning that turns on or off per request, under Apache 2.0.", "licence": "Apache-2.0", "run": "vllm serve poolside/Laguna-M.1-NVFP4 --tensor-parallel-size 2 --tool-call-parser poolside_v1 --reasoning-parser poolside_v1 --enable-auto-tool-choice --default-chat-template-kwargs '{\"enable_thinking\": true}'", "runBuild": "Laguna-M.1-NVFP4", "url": "/tools/laguna-m-1", "builds": [ { "name": "Laguna-M.1-NVFP4", "url": "https://huggingface.co/poolside/Laguna-M.1-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 139.3, "memoryGB": 192 }, { "name": "Laguna-M.1-FP8", "url": "https://huggingface.co/poolside/Laguna-M.1-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 238.6, "memoryGB": 512 }, { "name": "Laguna-M.1", "url": "https://huggingface.co/poolside/Laguna-M.1", "precision": "bf16", "format": "safetensors", "sizeGB": 451.6, "memoryGB": 1024, "runsOn": [ "vLLM", "SGLang", "Transformers" ] } ] }, { "slug": "qwen3-30b-a3b-instruct-2507", "name": "Qwen3-30B-A3B-Instruct-2507", "grade": "C", "score": 56.6, "tasks": [ "agents", "general" ], "params": "30.5B (3.3B active)", "context": 262144, "goodFor": "Mixture-of-experts text model with 3.3B of 30.5B parameters active per token, for general instruction following, tool calling and long documents without a thinking step.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3-30B-A3B-Instruct-2507 --max-model-len 262144 --enable-auto-tool-choice --tool-call-parser hermes", "runBuild": "Qwen3-30B-A3B-Instruct-2507", "url": "/tools/qwen3-30b-a3b-instruct-2507", "builds": [ { "name": "ollama qwen3:30b-a3b-instruct-2507-q4_K_M", "url": "https://ollama.com/library/qwen3:30b-a3b-instruct-2507-q4_K_M", "precision": "q4_k_m", "format": "ollama", "sizeGB": 19, "memoryGB": 24, "runsOn": [ "Ollama" ] }, { "name": "LM Studio qwen/qwen3-30b-a3b-2507 (Q4_K_M GGUF)", "url": "https://huggingface.co/lmstudio-community/Qwen3-30B-A3B-Instruct-2507-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 18.56, "memoryGB": 24, "runsOn": [ "LM Studio", "llama.cpp" ] }, { "name": "Qwen3-30B-A3B-Instruct-2507-FP8", "url": "https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 31.2, "memoryGB": 48 }, { "name": "Qwen3-30B-A3B-Instruct-2507", "url": "https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507", "precision": "bf16", "format": "safetensors", "sizeGB": 61.08, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "qwen3-vl-32b-instruct", "name": "Qwen3-VL-32B-Instruct", "grade": "C", "score": 56.4, "tasks": [ "agents", "general", "vision" ], "paramsB": 33.36, "params": "33.36B", "context": 262144, "goodFor": "The largest dense Qwen3-VL model, for images, documents and video with text out, about 21 GB at 4-bit, no thinking mode in this Instruct edition.", "licence": "Apache-2.0", "run": "llama-server -hf Qwen/Qwen3-VL-32B-Instruct-GGUF:Q4_K_M --jinja -c 32768", "runBuild": "Qwen3VL-32B-Instruct-Q4_K_M", "url": "/tools/qwen3-vl-32b-instruct", "builds": [ { "name": "Qwen3VL-32B-Instruct-Q4_K_M", "url": "https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct-GGUF/blob/main/Qwen3VL-32B-Instruct-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 20.53, "memoryGB": 32, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "ollama qwen3-vl:32b-instruct", "url": "https://ollama.com/library/qwen3-vl:32b-instruct", "precision": "q4_k_m", "format": "ollama", "sizeGB": 21, "memoryGB": 32, "runsOn": [ "Ollama" ] }, { "name": "Qwen3-VL-32B-Instruct-FP8", "url": "https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 35.53, "memoryGB": 48 }, { "name": "Qwen3VL-32B-Instruct-Q8_0", "url": "https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct-GGUF/blob/main/Qwen3VL-32B-Instruct-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 35.59, "memoryGB": 48, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Qwen3-VL-32B-Instruct", "url": "https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct", "precision": "bf16", "format": "safetensors", "sizeGB": 66.73, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] }, { "name": "Qwen3VL-32B-Instruct-F16", "url": "https://huggingface.co/Qwen/Qwen3-VL-32B-Instruct-GGUF/blob/main/Qwen3VL-32B-Instruct-F16-split-00001-of-00002.gguf", "precision": "f16", "format": "gguf", "sizeGB": 66.73, "memoryGB": 96, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] } ] }, { "slug": "bigbang-v1", "name": "BigBang-v1", "grade": "C", "score": 56.3, "tasks": [ "agents", "general", "vision" ], "activeB": 3, "params": "35B (3B active)", "context": 262144, "goodFor": "General-purpose mixture-of-experts model tuned for scientific research, search and coding tasks with thinking on, text and images in.", "licence": "Apache-2.0", "run": "vllm serve endless-frontier/BigBang-v1 --max-model-len 131072 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "BigBang-v1", "url": "/tools/bigbang-v1", "builds": [ { "name": "BigBang-v1", "url": "https://huggingface.co/endless-frontier/BigBang-v1", "precision": "bf16", "format": "safetensors", "sizeGB": 71.9, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)", "Transformers" ] } ] }, { "slug": "ling-3-0-tiny", "name": "Ling-3.0-tiny", "grade": "C", "score": 56.2, "tasks": [ "agents", "general" ], "params": "7.9B (1.3B active)", "context": 131072, "goodFor": "Small, fast hybrid reasoning model for local agents, tool calling and coding on a laptop, with thinking switchable per request.", "licence": "MIT", "run": "llama-server -hf inclusionAI/Ling-3.0-tiny-GGUF:Q4_K_M --jinja -c 32768", "runBuild": "Ling-3.0-tiny-Q4_K_M", "url": "/tools/ling-3-0-tiny", "builds": [ { "name": "Ling-3.0-tiny-int4", "url": "https://huggingface.co/inclusionAI/Ling-3.0-tiny-int4", "precision": "int4", "format": "safetensors", "sizeGB": 5.8, "memoryGB": 8 }, { "name": "Ling-3.0-tiny-Q6_K", "url": "https://huggingface.co/inclusionAI/Ling-3.0-tiny-GGUF/blob/main/Ling-3.0-tiny-Q6_K.gguf", "precision": "q6_k", "format": "gguf", "sizeGB": 6.5, "memoryGB": 8, "runsOn": [ "llama.cpp" ] }, { "name": "Ling-3.0-tiny-Q5_K_M", "url": "https://huggingface.co/inclusionAI/Ling-3.0-tiny-GGUF/blob/main/Ling-3.0-tiny-Q5_K_M.gguf", "precision": "q5_k_m", "format": "gguf", "sizeGB": 5.6, "memoryGB": 8, "runsOn": [ "llama.cpp" ] }, { "name": "Ling-3.0-tiny-Q4_K_M", "url": "https://huggingface.co/inclusionAI/Ling-3.0-tiny-GGUF/blob/main/Ling-3.0-tiny-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 4.8, "memoryGB": 8, "runsOn": [ "llama.cpp" ] }, { "name": "Ling-3.0-tiny-fp8", "url": "https://huggingface.co/inclusionAI/Ling-3.0-tiny-fp8", "precision": "fp8", "format": "safetensors", "sizeGB": 8.4, "memoryGB": 12 }, { "name": "Ling-3.0-tiny-Q8_0", "url": "https://huggingface.co/inclusionAI/Ling-3.0-tiny-GGUF/blob/main/Ling-3.0-tiny-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 8.4, "memoryGB": 12, "runsOn": [ "llama.cpp" ] }, { "name": "Ling-3.0-tiny", "url": "https://huggingface.co/inclusionAI/Ling-3.0-tiny", "precision": "bf16", "format": "safetensors", "sizeGB": 15.8, "memoryGB": 24 } ] }, { "slug": "lbs-27b", "name": "LBS-27B", "grade": "C", "score": 55.8, "tasks": [ "general", "vision" ], "paramsB": 27, "params": "27B", "context": 262144, "goodFor": "Finding the point to click on a screenshot, and answering questions about fine detail in large images, by zooming in over a few turns, on a machine with about 96 GB of memory and a client that runs the crops, if inclusionAI's own results hold up.", "licence": "Apache-2.0 (the card's metadata and licence section; the LICENSE file sits in the GitHub repository inclusionAI/UI-Venus on its LBS branch, not in the Hugging Face repository)", "run": "vllm serve inclusionAI/LBS-27B --served-model-name lbs --tensor-parallel-size 4 --max-model-len 65536", "runBuild": "LBS-27B (BF16)", "url": "/tools/lbs-27b", "builds": [ { "name": "LBS-27B (BF16)", "url": "https://huggingface.co/inclusionAI/LBS-27B", "precision": "bf16", "format": "safetensors", "sizeGB": 54.73, "memoryGB": 96, "runsOn": [ "vLLM", "SGLang", "Transformers" ] } ] }, { "slug": "qwen3-coder-30b-a3b-instruct", "name": "Qwen3-Coder-30B-A3B-Instruct", "grade": "C", "score": 55.7, "tasks": [ "agents", "general", "coding" ], "params": "30.5B (3.3B active)", "context": 262144, "goodFor": "Mixture-of-experts coding model with 3.3B of 30.5B parameters active per token, for agentic coding tools, repository-scale code reading, tool calls and fill-in-the-middle, with no thinking step.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3-Coder-30B-A3B-Instruct --max-model-len 262144 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "Qwen3-Coder-30B-A3B-Instruct", "url": "/tools/qwen3-coder-30b-a3b-instruct", "builds": [ { "name": "ollama qwen3-coder:30b", "url": "https://ollama.com/library/qwen3-coder:30b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 19, "memoryGB": 24, "runsOn": [ "Ollama" ] }, { "name": "LM Studio qwen/qwen3-coder-30b (Q4_K_M GGUF)", "url": "https://huggingface.co/lmstudio-community/Qwen3-Coder-30B-A3B-Instruct-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 18.63, "memoryGB": 24, "runsOn": [ "LM Studio", "llama.cpp" ] }, { "name": "Qwen3-Coder-30B-A3B-Instruct-FP8", "url": "https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 31.2, "memoryGB": 48 }, { "name": "Qwen3-Coder-30B-A3B-Instruct", "url": "https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct", "precision": "bf16", "format": "safetensors", "sizeGB": 61.08, "memoryGB": 96, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "fish-audio-drama-3", "name": "Fish Audio Drama 3", "grade": "C", "score": 55.6, "tasks": [ "tts" ], "params": "35.9B (35B backbone with about 3B active per token, per the card; 35,899,643,008 by the Hub's count)", "context": 16384, "goodFor": "Directed, expressive speech for multi-character scenes, long-form narration and dubbing, with voice cloning, revision by note and word-level editing, on an NVIDIA GPU.", "licence": "Fish Audio Community License 1.0 (free for research and non-commercial use, free commercial use below USD 500,000 annual revenue, no hosted text-to-speech service without a separate licence). The card's metadata still names fish-audio-research-license", "run": "git clone https://github.com/sgl-project/sglang-omni.git \u0026\u0026 cd sglang-omni \u0026\u0026 uv pip install --prerelease=allow -e \".[fishaudio]\" \u0026\u0026 hf download fishaudio/drama-3-int4 \u0026\u0026 sgl-omni serve --model-path fishaudio/drama-3-int4 --port 8000", "runBuild": "drama-3-int4", "url": "/tools/fish-audio-drama-3", "builds": [ { "name": "drama-3-int4", "url": "https://huggingface.co/fishaudio/drama-3-int4", "precision": "int4", "format": "safetensors", "sizeGB": 24, "memoryGB": 32 }, { "name": "drama-3", "url": "https://huggingface.co/fishaudio/drama-3", "precision": "fp8", "format": "safetensors", "sizeGB": 41.5, "memoryGB": 64 } ] }, { "slug": "drex-max-v1", "name": "Drex Max v1", "grade": "C", "score": 55.5, "tasks": [ "decide" ], "paramsB": 27, "params": "27B", "context": 262144, "goodFor": "Typed decisions with a probability per option over text, JSON and images, such as routing, triage and the next click in a computer-use loop, several questions per request, with no text generated.", "licence": "Nace.AI Open RAIL-M License (modified), version 1.0, October 2026, in the repository's LICENSE. Not an open-source licence. Use is free only for individuals and organisations under $1 million in revenue in the prior 12 months and under $1 million in total funding (personal and research use excepted), never by anyone with a product that competes with Nace's, decision models included, and under use restrictions that bind the model, its derivatives and its outputs. Built on Qwen/Qwen3.8-27B (Apache-2.0), whose licence text isn't carried in the repository", "run": "python -m drex_max --model weights/drex-max-v1 --request examples/request.json", "runBuild": "drex-max-v1", "url": "/tools/drex-max-v1", "builds": [ { "name": "drex-max-v1-affine-q4", "url": "https://huggingface.co/nace-ai/drex-max-v1-affine-q4", "precision": "affine 4-bit, group 64", "format": "safetensors (MLX)", "sizeGB": 14.5, "memoryGB": 24, "runsOn": [ "Kapitsa 0.2.25+", "MLX (drex_max_mlx.py)" ] }, { "name": "drex-max-v1-Q8_0", "url": "https://huggingface.co/nace-ai/drex-max-v1-Q8_0", "precision": "q8_0", "format": "gguf", "sizeGB": 27.3, "memoryGB": 48, "runsOn": [ "llama.cpp (Nace fork, branch drex-max-v1)", "Ollama (Nace fork, branch drex-max-v1)", "Kapitsa" ] }, { "name": "drex-max-v1", "url": "https://huggingface.co/nace-ai/drex-max-v1", "precision": "bf16", "format": "safetensors", "sizeGB": 52.2, "memoryGB": 64, "runsOn": [ "Transformers 5.17.0 with Nace's loader (CUDA)" ] } ] }, { "slug": "kronos", "name": "Kronos", "grade": "C", "score": 55.2, "tasks": [ "general" ], "params": "4.1M to 102.3M", "context": 512, "goodFor": "Forecasting the next bars of a financial candlestick series, or sampling synthetic ones, on any machine, from a short history of open, high, low, close, volume and amount.", "licence": "MIT", "run": "git clone https://github.com/shiyu-coder/Kronos \u0026\u0026 cd Kronos \u0026\u0026 pip install -r requirements.txt, then in Python: KronosPredictor(Kronos.from_pretrained(\"NeoQuasar/Kronos-small\"), KronosTokenizer.from_pretrained(\"NeoQuasar/Kronos-Tokenizer-base\"), max_context=512).predict(df=..., x_timestamp=..., y_timestamp=..., pred_len=120)", "runBuild": "Kronos-small", "url": "/tools/kronos", "builds": [ { "name": "Kronos-mini", "url": "https://huggingface.co/NeoQuasar/Kronos-mini", "precision": "fp32", "format": "safetensors", "sizeGB": 0.03, "memoryGB": 8, "runsOn": [ "PyTorch" ] }, { "name": "Kronos-small", "url": "https://huggingface.co/NeoQuasar/Kronos-small", "precision": "fp32", "format": "safetensors", "sizeGB": 0.11, "memoryGB": 8, "runsOn": [ "PyTorch" ] }, { "name": "Kronos-base", "url": "https://huggingface.co/NeoQuasar/Kronos-base", "precision": "fp32", "format": "safetensors", "sizeGB": 0.43, "memoryGB": 8, "runsOn": [ "PyTorch" ] } ] }, { "slug": "qwen3-vl-8b-instruct", "name": "Qwen3-VL-8B-Instruct", "grade": "C", "score": 54.9, "tasks": [ "agents", "general", "vision" ], "paramsB": 8.77, "params": "8.77B", "context": 262144, "goodFor": "Vision-language model for images, documents and video with text out, no thinking mode in this Instruct edition.", "licence": "Apache-2.0", "run": "llama-server -hf Qwen/Qwen3-VL-8B-Instruct-GGUF:Q4_K_M --jinja -c 32768", "runBuild": "Qwen3VL-8B-Instruct-Q4_K_M", "url": "/tools/qwen3-vl-8b-instruct", "builds": [ { "name": "Qwen3VL-8B-Instruct-Q4_K_M", "url": "https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct-GGUF/blob/main/Qwen3VL-8B-Instruct-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 5, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "ollama qwen3-vl:8b-instruct", "url": "https://ollama.com/library/qwen3-vl:8b-instruct", "precision": "q4_k_m", "format": "ollama", "sizeGB": 6.2, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "Qwen3VL-8B-Instruct-Q8_0", "url": "https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct-GGUF/blob/main/Qwen3VL-8B-Instruct-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 8.7, "memoryGB": 12, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Qwen3-VL-8B-Instruct-FP8", "url": "https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 10.6, "memoryGB": 16 }, { "name": "Qwen3-VL-8B-Instruct", "url": "https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct", "precision": "bf16", "format": "safetensors", "sizeGB": 17.5, "memoryGB": 24, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] }, { "name": "Qwen3VL-8B-Instruct-F16", "url": "https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct-GGUF/blob/main/Qwen3VL-8B-Instruct-F16.gguf", "precision": "f16", "format": "gguf", "sizeGB": 16.4, "memoryGB": 24, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] } ] }, { "slug": "ming-3-0-asr-tiny", "name": "Ming-3.0-asr-tiny", "grade": "C", "score": 54.8, "tasks": [ "stt" ], "paramsB": 8.87, "params": "8.87B", "goodFor": "Speech recognition for Chinese, English and Chinese dialects, taking hotwords or dialogue context as hints.", "licence": "MIT", "run": "hf download inclusionAI/Ming-3.0-asr-tiny --local-dir Ming-3.0-asr-tiny \u0026\u0026 cd Ming-3.0-asr-tiny \u0026\u0026 pip install -r requirements.txt \u0026\u0026 python examples/transcribe.py --model . --audio /path/to/speech.wav --language-hint en", "runBuild": "Ming-3.0-asr-tiny", "url": "/tools/ming-3-0-asr-tiny", "builds": [ { "name": "Ming-3.0-asr-tiny", "url": "https://huggingface.co/inclusionAI/Ming-3.0-asr-tiny", "precision": "bf16", "format": "safetensors", "sizeGB": 17.7, "memoryGB": 24 } ] }, { "slug": "mistral-large-4", "name": "Mistral Large 4", "grade": "C", "score": 54.6, "tasks": [ "agents", "general", "vision" ], "params": "~1T total (1.05T per Mistral's docs), 52B active", "context": 1048576, "goodFor": "Frontier-sized multimodal model for coding agents, defensive security work and long documents, with tool calling, strict JSON Schema output and three reasoning efforts, for operators with a multi-GPU server.", "licence": "Mistral Ecosystem Open-Weight License (modified MIT)", "run": "docker run --runtime nvidia --gpus all -v ~/.cache/huggingface:/root/.cache/huggingface -p 8000:8000 --ipc=host vllm/vllm-openai:mistral-large-4 --model mistralai/Mistral-Large-4-1T-A52B-NVFP4 --tensor-parallel-size 8 --kv-cache-dtype fp8_ds_mla --tool-call-parser mistral --reasoning-parser mistral --enable-auto-tool-choice --gpu-memory-utilization 0.85", "runBuild": "Mistral-Large-4-1T-A52B-NVFP4", "url": "/tools/mistral-large-4", "builds": [ { "name": "Mistral-Large-4-1T-A52B-NVFP4", "url": "https://huggingface.co/mistralai/Mistral-Large-4-1T-A52B-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 610, "memoryGB": 1024, "runsOn": [ "vLLM (mistral-large-4 image)", "SGLang (mistral-large-4 image)" ] }, { "name": "ggml-org Mistral-Large-4-1T-A52B-GGUF Q4_K_M", "url": "https://huggingface.co/ggml-org/Mistral-Large-4-1T-A52B-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 634.8, "memoryGB": 1024, "runsOn": [ "llama.cpp" ] }, { "name": "Mistral-Large-4-1T-A52B", "url": "https://huggingface.co/mistralai/Mistral-Large-4-1T-A52B", "precision": "fp8", "format": "safetensors", "sizeGB": 1060.6, "memoryGB": 2048, "runsOn": [ "Transformers", "vLLM (mistral-large-4 image)", "SGLang (mistral-large-4 image)" ] } ] }, { "slug": "ming-3-0-tts-tiny", "name": "Ming-3.0-tts-tiny", "grade": "C", "score": 54.5, "tasks": [ "tts" ], "paramsB": 7.9, "activeB": 1.3, "params": "7.9B", "goodFor": "Text to speech with voice cloning, voice design and controlled speech in seven languages, on a mixture-of-experts backbone.", "licence": "MIT", "run": "git clone https://github.com/inclusionAI/Ming-3.0-tts.git \u0026\u0026 cd Ming-3.0-tts \u0026\u0026 pip install -r requirements-tiny.txt \u0026\u0026 bash examples/one_click/run_seed.sh", "runBuild": "Ming-3.0-tts-tiny", "url": "/tools/ming-3-0-tts-tiny", "builds": [ { "name": "Ming-3.0-tts-tiny", "url": "https://huggingface.co/inclusionAI/Ming-3.0-tts-tiny", "precision": "bf16", "format": "safetensors", "sizeGB": 18, "memoryGB": 24 } ] }, { "slug": "qwen3-4b-instruct-2507", "name": "Qwen3-4B-Instruct-2507", "grade": "C", "score": 54.5, "tasks": [ "agents", "general" ], "paramsB": 4, "params": "4B", "context": 262144, "goodFor": "Small dense text model for local and on-device use, instruction following and tool calling with a long context and no thinking step.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3-4B-Instruct-2507 --max-model-len 262144 --enable-auto-tool-choice --tool-call-parser hermes", "runBuild": "Qwen3-4B-Instruct-2507", "url": "/tools/qwen3-4b-instruct-2507", "builds": [ { "name": "Qwen3-4B-Instruct-2507-FP8", "url": "https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 5.21, "memoryGB": 8 }, { "name": "ollama qwen3:4b-instruct-2507-q4_K_M", "url": "https://ollama.com/library/qwen3:4b-instruct-2507-q4_K_M", "precision": "q4_k_m", "format": "ollama", "sizeGB": 2.5, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "LM Studio qwen/qwen3-4b-2507 (Q4_K_M GGUF)", "url": "https://huggingface.co/lmstudio-community/Qwen3-4B-Instruct-2507-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 2.5, "memoryGB": 8, "runsOn": [ "LM Studio", "llama.cpp" ] }, { "name": "Qwen3-4B-Instruct-2507", "url": "https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507", "precision": "bf16", "format": "safetensors", "sizeGB": 8.06, "memoryGB": 12, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] } ] }, { "slug": "minimax-m3-1-flash", "name": "MiniMax-M3.1-Flash", "grade": "C", "score": 54.3, "tasks": [ "agents", "general", "coding", "vision" ], "paramsB": 428, "params": "428B", "context": 1048576, "goodFor": "Mixture-of-experts model for coding, agent harnesses, computer use and security work, with images and video in, about 1M tokens of context and parallel tool calls, on a 4 to 8 GPU server.", "licence": "MiniMax Community License (custom; commercial use needs a \"Built with MiniMax M3.1-Flash\" line and a one-time notice to MiniMax, and MiniMax's written authorisation above $20M yearly revenue)", "run": "docker run --gpus all --ipc=host -p 8000:8000 -v ~/.cache/huggingface:/root/.cache/huggingface vllm/vllm-openai:minimax-m31-flash MiniMaxAI/MiniMax-M3.1-Flash --tensor-parallel-size 8 --block-size 128 --disable-hybrid-kv-cache-manager --kv-cache-dtype fp8 --attention-config '{\"indexer_kv_dtype\": \"fp8\"}' --moe-backend marlin --max-model-len 262144 --reasoning-parser minimax_m3 --enable-auto-tool-choice --tool-call-parser minimax_m3", "runBuild": "MiniMax-M3.1-Flash", "url": "/tools/minimax-m3-1-flash", "builds": [ { "name": "MiniMax-M3.1-Flash", "url": "https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash", "precision": "nvfp4", "format": "safetensors", "sizeGB": 250.12, "memoryGB": 512, "runsOn": [ "vLLM", "SGLang" ] } ] }, { "slug": "unlimited-ocr-2", "name": "Baidu Unlimited-OCR-2", "grade": "C", "score": 54.2, "tasks": [ "general", "vision", "docs" ], "params": "3.3B (3B-A0.5B mixture-of-experts decoder per the report)", "context": 131072, "goodFor": "Parsing long scanned or rendered documents, tens to hundreds of pages, into ordered layout blocks with boxes and HTML tables in one request, on an NVIDIA GPU, if Baidu's own results hold up.", "licence": "MIT (Hub metadata; the Hugging Face repository has no LICENSE file, and the GitHub code repository's LICENSE is the MIT text, copyright 2026 Baidu)", "run": "hf download baidu/Unlimited-OCR-2 --local-dir uocr2 \u0026\u0026 uv venv --python 3.10 \u0026\u0026 . .venv/bin/activate \u0026\u0026 uv pip install uocr2/wheel/sglang-0.5.20.dev13+g75b9b06b02-py3-none-any.whl pymupdf==1.27.2.2 fastokens==0.3.2 \u0026\u0026 SGLANG_UOCR_EANABLE_DOC_LIMIT=1 SGLANG_UOCR_DOC_GUNDAM_PAGE_WINDOW=10 SGLANG_UOCR_DOC_BASE_PAGE_WINDOW=10 SGLANG_UOCR_MAX_PAGE_TOKENS=20000 python -m sglang.launch_server --model-path baidu/Unlimited-OCR-2 --served-model-name Unlimited-OCR-2 --chat-template unlimited-ocr --tokenizer-backend fastokens --mem-fraction-static 0.7 --max-prefill-tokens 2048 --enable-custom-logit-processor --enable-streaming-session --allow-auto-truncate --disable-radix-cache --port 10000", "runBuild": "Unlimited-OCR-2 (BF16 safetensors)", "url": "/tools/unlimited-ocr-2", "builds": [ { "name": "Unlimited-OCR-2 (BF16 safetensors)", "url": "https://huggingface.co/baidu/Unlimited-OCR-2", "precision": "bf16", "format": "safetensors", "sizeGB": 6.69, "memoryGB": 12, "runsOn": [ "Transformers (remote code)", "SGLang (Baidu's customised wheel)" ] } ] }, { "slug": "swift-1-5-qwen3-8-flash-next", "name": "Swift 1.5 Qwen3.8-Flash-Next", "grade": "C", "score": 54.2, "tasks": [ "agents", "general", "vision" ], "paramsB": 125, "params": "125B", "context": 262144, "goodFor": "A Qwen3.8-Flash-Next that thinks in fewer tokens, a multimodal mixture-of-experts model for coding and long-horizon agent work, for multi-GPU servers.", "licence": "Swift Open License v1.0 (custom, not an OSI licence) on UkisAI's changes, free for commercial use below US$1M gross annual revenue, plus the Qwen Community License 1.0 on the base, which needs a separate licence from Qwen for a Model as a Service or AI Work Assistant business and the model name shown above 100M monthly users or US$20M monthly revenue", "run": "vllm serve ukisai/Swift1.5-Qwen3.8-Flash-Next --tensor-parallel-size 8 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "Swift1.5-Qwen3.8-Flash-Next", "url": "/tools/swift-1-5-qwen3-8-flash-next", "builds": [ { "name": "Swift-1.5-Qwen3.8-Flash-Next-GGUF IQ4_XS", "url": "https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-Flash-Next-GGUF", "precision": "iq4_xs", "format": "gguf", "sizeGB": 98.6, "memoryGB": 128, "runsOn": [ "llama.cpp" ] }, { "name": "Swift-1.5-Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S", "url": "https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-Flash-Next-GSQ-RCO-GGUF", "precision": "iq3_s", "format": "gguf", "sizeGB": 84.7, "memoryGB": 128, "runsOn": [ "llama.cpp" ] }, { "name": "Swift-1.5-Qwen3.8-Flash-Next-GGUF Q4_K_M", "url": "https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-Flash-Next-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 120.5, "memoryGB": 192, "runsOn": [ "llama.cpp" ] }, { "name": "Swift-1.5-Qwen3.8-Flash-Next-W4A16-AWQ", "url": "https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-Flash-Next-W4A16-AWQ", "precision": "int4", "format": "safetensors", "sizeGB": 177.3, "memoryGB": 256, "runsOn": [ "vLLM" ] }, { "name": "Swift-1.5-Qwen3.8-Flash-Next-NVFP4", "url": "https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-Flash-Next-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 186.5, "memoryGB": 256 }, { "name": "Swift1.5-Qwen3.8-Flash-Next", "url": "https://huggingface.co/ukisai/Swift1.5-Qwen3.8-Flash-Next", "precision": "bf16", "format": "safetensors", "sizeGB": 360, "memoryGB": 512 } ] }, { "slug": "elephant-1", "name": "elephant-1", "grade": "D", "score": 53.6, "tasks": [ "decide" ], "paramsB": 9, "params": "9B", "context": 16384, "goodFor": "Typed decisions with calibrated probabilities over text and images, several choice, yes or no and score questions about one state in one prefill, for routing, triage, guardrails and judging.", "licence": "MIT, named only in the card's metadata (`license: mit`) since 21 October 2026, with no LICENSE file. Until then the card said proprietary, all rights reserved, access for Decision Index evaluation only. The base Qwen/Qwen3.5-9B is Apache-2.0", "run": "hf download baselabs/elephant-1 --local-dir elephant-1 \u0026\u0026 bash elephant-1/setup.sh, then qev's DecisionEngine(\"baselabs/elephant-1\", device=\"cuda:0\").decide(SystemOneRequest(state=..., questions=...))", "runBuild": "elephant-1", "url": "/tools/elephant-1", "builds": [ { "name": "elephant-1", "url": "https://huggingface.co/baselabs/elephant-1", "precision": "bf16", "format": "safetensors", "sizeGB": 18.8, "memoryGB": 24, "runsOn": [ "OneJev qev with serve.patch" ] } ] }, { "slug": "ripple-65b-preview", "name": "Ripple-65B Preview", "grade": "D", "score": 53.3, "tasks": [ "agents", "general", "coding" ], "paramsB": 65, "params": "65B", "context": 1048576, "goodFor": "A dense 65B model post-trained for agentic coding and terminal work, with several tool calls a turn and a long context, for teams with Blackwell GPUs or a DGX Spark.", "licence": "Apache-2.0", "run": "CUDA_VISIBLE_DEVICES=0,1 uv run --script serve.py --model-path /path/to/Ripple-65B-Preview --speculative-draft-model-path /path/to/Ripple-65B-Preview-DFlash2", "runBuild": "Ripple-65B-Preview", "url": "/tools/ripple-65b-preview", "builds": [ { "name": "Ripple-65B-Preview-NVFP4", "url": "https://huggingface.co/ricursiveAI/Ripple-65B-Preview-NVFP4", "precision": "nvfp4", "format": "safetensors", "sizeGB": 57.2, "memoryGB": 69 }, { "name": "Ripple-65B-Preview", "url": "https://huggingface.co/ricursiveAI/Ripple-65B-Preview", "precision": "bf16", "format": "safetensors", "sizeGB": 129.6, "memoryGB": 160 } ] }, { "slug": "internvl3-5-gpt-oss-20b-a4b-preview", "name": "InternVL3.5-GPT-OSS-20B-A4B-Preview", "grade": "D", "score": 53.2, "tasks": [ "general", "vision" ], "paramsB": 21.2, "activeB": 4, "params": "21.2B", "context": 32768, "goodFor": "Vision-language preview for images, documents, charts and video with text out and an optional thinking mode, on the gpt-oss-20b mixture-of-experts backbone.", "licence": "Apache-2.0; a merge of openai/gpt-oss-20b (Apache 2.0 with OpenAI's usage policy) and OpenGVLab/InternViT-300M-448px-V2_5 (MIT)", "run": "vllm serve OpenGVLab/InternVL3_5-GPT-OSS-20B-A4B-Preview-HF --max-model-len 32768", "runBuild": "InternVL3_5-GPT-OSS-20B-A4B-Preview-HF", "url": "/tools/internvl3-5-gpt-oss-20b-a4b-preview", "builds": [ { "name": "InternVL3_5-GPT-OSS-20B-A4B-Preview-HF", "url": "https://huggingface.co/OpenGVLab/InternVL3_5-GPT-OSS-20B-A4B-Preview-HF", "precision": "bf16", "format": "safetensors", "sizeGB": 42.5, "memoryGB": 64, "runsOn": [ "Transformers", "vLLM" ] }, { "name": "InternVL3_5-GPT-OSS-20B-A4B-Preview", "url": "https://huggingface.co/OpenGVLab/InternVL3_5-GPT-OSS-20B-A4B-Preview", "precision": "bf16", "format": "safetensors", "sizeGB": 42.5, "memoryGB": 64, "runsOn": [ "SGLang", "llama.cpp (after conversion)", "Ollama (through its llama.cpp pin, after conversion)", "Transformers (remote code)" ] } ] }, { "slug": "qwen3-vl-4b-instruct", "name": "Qwen3-VL-4B-Instruct", "grade": "D", "score": 52.6, "tasks": [ "agents", "general", "vision" ], "paramsB": 4.44, "params": "4.44B", "context": 262144, "goodFor": "Small vision-language model for images, documents and video with text out, fits in 8 GB at 4-bit, no thinking mode in this Instruct edition.", "licence": "Apache-2.0", "run": "llama-server -hf Qwen/Qwen3-VL-4B-Instruct-GGUF:Q4_K_M --jinja -c 32768", "runBuild": "Qwen3VL-4B-Instruct-Q4_K_M", "url": "/tools/qwen3-vl-4b-instruct", "builds": [ { "name": "Qwen3-VL-4B-Instruct-FP8", "url": "https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct-FP8", "precision": "fp8", "format": "safetensors", "sizeGB": 6.04, "memoryGB": 8 }, { "name": "Qwen3VL-4B-Instruct-Q8_0", "url": "https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct-GGUF/blob/main/Qwen3VL-4B-Instruct-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 4.73, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "Qwen3VL-4B-Instruct-Q4_K_M", "url": "https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct-GGUF/blob/main/Qwen3VL-4B-Instruct-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 2.95, "memoryGB": 8, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] }, { "name": "ollama qwen3-vl:4b-instruct", "url": "https://ollama.com/library/qwen3-vl:4b-instruct", "precision": "q4_k_m", "format": "ollama", "sizeGB": 3.3, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "Qwen3-VL-4B-Instruct", "url": "https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct", "precision": "bf16", "format": "safetensors", "sizeGB": 8.89, "memoryGB": 12, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] }, { "name": "Qwen3VL-4B-Instruct-F16", "url": "https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct-GGUF/blob/main/Qwen3VL-4B-Instruct-F16.gguf", "precision": "f16", "format": "gguf", "sizeGB": 8.89, "memoryGB": 12, "runsOn": [ "llama.cpp (Vulkan)", "llama.cpp (ROCm)" ] } ] }, { "slug": "fish-audio-s2-pro", "name": "Fish Audio S2 Pro", "grade": "D", "score": 52.5, "tasks": [ "tts" ], "params": "4.56B (slow AR 4B plus fast AR 0.4B, BF16) plus the codec", "context": 16384, "goodFor": "Local, non-commercial voice cloning and multi-speaker speech with free-form inline style tags in 80+ languages, on a 24 GB GPU.", "licence": "Fish Audio Research License (weights; research and non-commercial use only); the fish-speech code that runs it is under the Fish Audio Community License 1.0 since 13 October 2026", "run": "python tools/api_server.py --listen 127.0.0.1:8080 --llama-checkpoint-path checkpoints/s2-pro --decoder-checkpoint-path checkpoints/s2-pro/codec.pth --decoder-config-name modded_dac_vq", "runBuild": "s2-pro", "url": "/tools/fish-audio-s2-pro", "builds": [ { "name": "s2-pro", "url": "https://huggingface.co/fishaudio/s2-pro", "precision": "bf16", "format": "safetensors", "sizeGB": 11.01, "memoryGB": 16 } ] }, { "slug": "olmo-3-7b-think", "name": "Olmo 3 7B Think", "grade": "D", "score": 52.3, "tasks": [ "general" ], "paramsB": 7.3, "params": "7.3B", "context": 65536, "goodFor": "Fully open 7B reasoning model for maths, code and step-by-step answers in English, with published training data and checkpoints for research, always thinking, not trained for tool calls.", "licence": "Apache-2.0, with Ai2's Responsible Use Guidelines named on the card", "run": "vllm serve allenai/Olmo-3-7B-Think --max-model-len 65536", "runBuild": "Olmo-3-7B-Think", "url": "/tools/olmo-3-7b-think", "builds": [ { "name": "ollama olmo-3:7b-think-q4_K_M", "url": "https://ollama.com/library/olmo-3:7b-think-q4_K_M", "precision": "q4_k_m", "format": "ollama", "sizeGB": 4.5, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "LM Studio allenai/olmo-3-7b-think (Q4_K_M GGUF)", "url": "https://huggingface.co/lmstudio-community/Olmo-3-7B-Think-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 4.5, "memoryGB": 8, "runsOn": [ "LM Studio", "llama.cpp" ] }, { "name": "Olmo-3-7B-Think", "url": "https://huggingface.co/allenai/Olmo-3-7B-Think", "precision": "bf16", "format": "safetensors", "sizeGB": 14.6, "memoryGB": 24, "runsOn": [ "vLLM", "Transformers" ] } ] }, { "slug": "ming-3-0-audio-tiny", "name": "Ming-3.0-audio-tiny", "grade": "D", "score": 52.1, "tasks": [ "stt" ], "paramsB": 9.3, "activeB": 2.7, "params": "9.3B", "context": 131072, "goodFor": "Small audio-language model for Chinese and English speech recognition, audio understanding and spoken replies.", "licence": "MIT", "run": "sgl-omni serve --model-path inclusionAI/Ming-3.0-audio-tiny --talker.gpu 0", "runBuild": "Ming-3.0-audio-tiny", "url": "/tools/ming-3-0-audio-tiny", "builds": [ { "name": "Ming-3.0-audio-tiny", "url": "https://huggingface.co/inclusionAI/Ming-3.0-audio-tiny", "precision": "bf16", "format": "safetensors", "sizeGB": 20.7, "memoryGB": 32 } ] }, { "slug": "audio8-tts-realistic-0-6b", "name": "Audio8-TTS-Realistic 0.6B", "grade": "D", "score": 51.5, "tasks": [ "tts" ], "params": "0.6B (601,159,424 BF16 parameters by the Hub's count, without the bundled codec)", "context": 2048, "goodFor": "Zero-shot voice cloning from one clip and its transcript, or speech without a reference, in 23 languages, on a small CUDA GPU, under Apache 2.0.", "licence": "Apache-2.0", "run": "git clone https://github.com/Edge0-AI/Audio8_TTS.git \u0026\u0026 cd Audio8_TTS \u0026\u0026 pip install -r requirements.txt \u0026\u0026 python audio8_tts_infer.py --model Edge0/Audio8-TTS-Realistic-Air --text \"Welcome to Audio8 TTS.\" --output outputs/realistic.wav", "runBuild": "Audio8-TTS-Realistic-Air", "url": "/tools/audio8-tts-realistic-0-6b", "builds": [ { "name": "Audio8-TTS-Realistic-Air", "url": "https://huggingface.co/Edge0/Audio8-TTS-Realistic-Air", "precision": "bf16", "format": "safetensors", "sizeGB": 2.57, "memoryGB": 8 } ] }, { "slug": "diffusiongemma-26b-a4b-it", "name": "DiffusionGemma 26B A4B IT", "grade": "D", "score": 51.3, "tasks": [ "agents", "general", "vision" ], "paramsB": 25.2, "activeB": 3.8, "params": "25.2B", "context": 262144, "goodFor": "Fast local text generation from text, image and video prompts at low concurrency, with optional thinking and tool calls, trading some quality against Gemma 4 26B A4B for speed.", "licence": "Apache-2.0", "run": "sglang serve --model-path google/diffusiongemma-26B-A4B-it --dllm-algorithm Gemma4Renoise --trust-remote-code --port 30000", "runBuild": "diffusiongemma-26B-A4B-it", "url": "/tools/diffusiongemma-26b-a4b-it", "builds": [ { "name": "diffusiongemma-26B-A4B-it", "url": "https://huggingface.co/google/diffusiongemma-26B-A4B-it", "precision": "bf16", "format": "safetensors", "sizeGB": 51.7, "memoryGB": 64, "runsOn": [ "Transformers", "SGLang", "vLLM" ] } ] }, { "slug": "qwen3-5-2b", "name": "Qwen3.5-2B", "grade": "D", "score": 50.5, "tasks": [ "agents", "general", "vision" ], "paramsB": 2, "params": "2B", "context": 262144, "goodFor": "Small vision-language model for prototyping, fine-tuning and light local agent steps: text, images and video in, tool calling, thinking off by default.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3.5-2B --port 8000 --max-model-len 262144 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "Qwen3.5-2B", "url": "/tools/qwen3-5-2b", "builds": [ { "name": "Qwen3.5-2B", "url": "https://huggingface.co/Qwen/Qwen3.5-2B", "precision": "bf16", "format": "safetensors", "sizeGB": 4.6, "memoryGB": 8, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] }, { "name": "ollama qwen3.5:2b-q4_K_M", "url": "https://ollama.com/library/qwen3.5:2b-q4_K_M", "precision": "q4_k_m", "format": "ollama", "sizeGB": 1.9, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "LM Studio qwen/qwen3.5-2b (Q4_K_M GGUF)", "url": "https://huggingface.co/lmstudio-community/Qwen3.5-2B-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 1.9, "memoryGB": 8, "runsOn": [ "LM Studio", "llama.cpp" ] } ] }, { "slug": "lfm2-5-2-6b", "name": "LFM2.5-2.6B", "grade": "D", "score": 50.3, "tasks": [ "agents", "general" ], "paramsB": 2.69, "params": "2.69B", "context": 131072, "goodFor": "Small on-device agent model for tool use, data extraction, RAG and long-context work on a laptop or phone, always thinking, not for agentic coding or knowledge-heavy tasks.", "licence": "LFM Open License v1.0 (free commercial use below $10M annual revenue)", "run": "llama-server -hf LiquidAI/LFM2.5-2.6B-GGUF:Q4_0 --jinja -c 131072 --temp 0.1 --top-k 50 --repeat-penalty 1.1", "runBuild": "LFM2.5-2.6B-GGUF Q4_0", "url": "/tools/lfm2-5-2-6b", "builds": [ { "name": "LFM2.5-2.6B", "url": "https://huggingface.co/LiquidAI/LFM2.5-2.6B", "precision": "bf16", "format": "safetensors", "sizeGB": 5.4, "memoryGB": 8, "runsOn": [ "vLLM", "SGLang", "Transformers" ] }, { "name": "LFM2.5-2.6B-GGUF Q4_0", "url": "https://huggingface.co/LiquidAI/LFM2.5-2.6B-GGUF", "precision": "q4_0", "format": "gguf", "sizeGB": 1.6, "memoryGB": 8, "runsOn": [ "llama.cpp", "LM Studio" ] }, { "name": "LFM2.5-2.6B-MLX 4bit", "url": "https://huggingface.co/LiquidAI/LFM2.5-2.6B-MLX", "precision": "4bit", "format": "mlx", "sizeGB": 1.6, "memoryGB": 8, "runsOn": [ "MLX" ] }, { "name": "ollama LiquidAI/lfm2.5-2.6b:q4_k_m", "url": "https://ollama.com/LiquidAI/lfm2.5-2.6b", "precision": "q4_k_m", "format": "ollama", "sizeGB": 1.7, "memoryGB": 8, "runsOn": [ "Ollama" ] } ] }, { "slug": "embeddinggemma-300m", "name": "EmbeddingGemma 300M", "grade": "D", "score": 48.9, "tasks": [ "embed" ], "paramsB": 0.308, "params": "308M", "context": 2048, "goodFor": "Small multilingual text embeddings for search, retrieval, classification and clustering on modest hardware, with task prompts and truncatable dimensions.", "licence": "Gemma Terms of Use", "run": "ollama pull embeddinggemma:300m-qat-q4_0", "runBuild": "ollama embeddinggemma:300m-qat-q4_0", "url": "/tools/embeddinggemma-300m", "builds": [ { "name": "embeddinggemma-300m", "url": "https://huggingface.co/google/embeddinggemma-300m", "precision": "f32", "format": "safetensors", "sizeGB": 1.21, "memoryGB": 8 }, { "name": "ollama embeddinggemma:300m-qat-q4_0", "url": "https://ollama.com/library/embeddinggemma:300m-qat-q4_0", "precision": "q4_0", "format": "ollama", "sizeGB": 0.24, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "lmstudio-community embeddinggemma-300m-qat-Q4_0", "url": "https://huggingface.co/lmstudio-community/embeddinggemma-300m-qat-GGUF", "precision": "q4_0", "format": "gguf", "sizeGB": 0.23, "memoryGB": 8, "runsOn": [ "LM Studio", "llama.cpp" ] } ] }, { "slug": "jina-embeddings-v5-omni-small", "name": "jina-embeddings-v5-omni-small", "grade": "D", "score": 48.8, "tasks": [ "embed" ], "paramsB": 1.63, "params": "1.63B", "context": 32768, "goodFor": "One vector space for text, images, video, audio and PDFs, compatible with an index built by jina-embeddings-v5-text-small, for non-commercial use.", "licence": "CC-BY-NC-4.0", "run": "vllm serve jinaai/jina-embeddings-v5-omni-small --trust-remote-code --hf-overrides '{\"task\": \"retrieval\"}'", "runBuild": "jina-embeddings-v5-omni-small", "url": "/tools/jina-embeddings-v5-omni-small", "builds": [ { "name": "jina-embeddings-v5-omni-small", "url": "https://huggingface.co/jinaai/jina-embeddings-v5-omni-small", "precision": "bf16", "format": "safetensors", "sizeGB": 3.43, "memoryGB": 8, "runsOn": [ "Transformers (remote code)", "Sentence Transformers (remote code)", "vLLM 0.20.1 (remote code)" ] }, { "name": "jina-embeddings-v5-omni-small-retrieval-Q4_K_M", "url": "https://huggingface.co/jinaai/jina-embeddings-v5-omni-small-retrieval-GGUF/blob/main/jina-embeddings-v5-omni-small-retrieval-Q4_K_M.gguf", "precision": "q4_k_m", "format": "gguf", "sizeGB": 0.4, "memoryGB": 8, "runsOn": [ "Jina's llama.cpp fork (feat-v5-omni)" ] }, { "name": "jina-embeddings-v5-omni-small-retrieval-Q8_0", "url": "https://huggingface.co/jinaai/jina-embeddings-v5-omni-small-retrieval-GGUF/blob/main/jina-embeddings-v5-omni-small-retrieval-Q8_0.gguf", "precision": "q8_0", "format": "gguf", "sizeGB": 0.64, "memoryGB": 8, "runsOn": [ "Jina's llama.cpp fork (feat-v5-omni)" ] }, { "name": "jina-embeddings-v5-omni-small-mlx", "url": "https://huggingface.co/jinaai/jina-embeddings-v5-omni-small-mlx", "precision": "bf16", "format": "mlx", "sizeGB": 3.29, "memoryGB": 8 } ] }, { "slug": "mammoth-1", "name": "mammoth-1", "grade": "D", "score": 48.6, "tasks": [ "decide" ], "paramsB": 27, "params": "27B", "context": 16384, "goodFor": "Typed decisions with a probability per option, several questions about one state answered together from one prefill, for routing, triage, classification, scoring and judging.", "licence": "MIT, named only in the card's metadata (`license: mit`) with no LICENSE file. Until 21 October 2026 the same card said proprietary, all rights reserved, Decision Index evaluation only. The base, Qwen/Qwen3.5-27B, is Apache-2.0", "run": "hf download baselabs/mammoth-1 --local-dir mammoth-1 \u0026\u0026 bash mammoth-1/setup.sh, then qev's DecisionEngine(\"baselabs/mammoth-1\", device=\"cuda:0\").decide(SystemOneRequest(state=..., questions=...))", "runBuild": "mammoth-1", "url": "/tools/mammoth-1", "builds": [ { "name": "mammoth-1", "url": "https://huggingface.co/baselabs/mammoth-1", "precision": "bf16", "format": "safetensors", "sizeGB": 54.7, "memoryGB": 96, "runsOn": [ "OneJev qev @81ce62f + serve.patch" ] } ] }, { "slug": "underdog-bark-3", "name": "Underdog Bark 3.0", "grade": "D", "score": 48.2, "tasks": [ "stt" ], "paramsB": 0.6, "params": "0.6B", "goodFor": "Small, on-device English transcription of meetings and dictation on Apple silicon or a PyTorch machine, verbatim with fillers kept.", "licence": "CC-BY-4.0", "run": "python bark_loader.py model.bark recording.wav", "runBuild": "model.bark", "url": "/tools/underdog-bark-3", "builds": [ { "name": "model.bark", "url": "https://huggingface.co/ConwayResearch/Underdog-Bark-3.0/blob/main/model.bark", "precision": "low-bit (entropy-coded per-row levels for encoder linear weights, 5 or 6 bit integers for other matrices, fp16 small tensors)", "format": "bark", "sizeGB": 0.16, "memoryGB": 8, "runsOn": [ "Transformers", "mlx-audio" ] }, { "name": "Core ML (Apple GPU)", "url": "https://huggingface.co/ConwayResearch/Underdog-Bark-3.0/tree/main/coreml/gpu", "precision": "the same quantised values as model.bark, FP16 encoder arithmetic", "format": "coreml", "sizeGB": 0.2, "memoryGB": 8, "runsOn": [ "Core ML" ] }, { "name": "Core ML (Apple Neural Engine)", "url": "https://huggingface.co/ConwayResearch/Underdog-Bark-3.0/tree/main/coreml/ane", "precision": "the same quantised values as model.bark, FP16 encoder arithmetic", "format": "coreml", "sizeGB": 0.2, "memoryGB": 8, "runsOn": [ "Core ML" ] } ] }, { "slug": "qwen3-5-0-8b", "name": "Qwen3.5-0.8B", "grade": "D", "score": 46.7, "tasks": [ "agents", "general", "vision" ], "paramsB": 0.8, "params": "0.8B", "context": 262144, "goodFor": "Tiny vision-language model for prototyping, fine-tuning and light local steps: text, images and video in, tool calling, thinking off by default.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3.5-0.8B --port 8000 --max-model-len 262144 --enable-auto-tool-choice --tool-call-parser qwen3_coder", "runBuild": "Qwen3.5-0.8B", "url": "/tools/qwen3-5-0-8b", "builds": [ { "name": "Qwen3.5-0.8B", "url": "https://huggingface.co/Qwen/Qwen3.5-0.8B", "precision": "bf16", "format": "safetensors", "sizeGB": 1.8, "memoryGB": 8, "runsOn": [ "vLLM (ROCm)", "SGLang (ROCm)" ] }, { "name": "ollama qwen3.5:0.8b-q8_0", "url": "https://ollama.com/library/qwen3.5:0.8b-q8_0", "precision": "q8_0", "format": "ollama", "sizeGB": 1, "memoryGB": 8, "runsOn": [ "Ollama" ] }, { "name": "LM Studio qwen/qwen3.5-0.8b (Q4_K_M GGUF)", "url": "https://huggingface.co/lmstudio-community/Qwen3.5-0.8B-GGUF", "precision": "q4_k_m", "format": "gguf", "sizeGB": 0.7, "memoryGB": 8, "runsOn": [ "LM Studio", "llama.cpp" ] } ] }, { "slug": "ming-3-0-audio-flash", "name": "Ming-3.0-audio-flash", "grade": "E", "score": 42.5, "tasks": [ "stt" ], "paramsB": 126, "activeB": 6.5, "params": "126B", "context": 131072, "goodFor": "Audio-language model for Chinese and English: speech to text, questions about audio and spoken replies.", "licence": "MIT", "run": "hf download inclusionAI/Ming-3.0-audio-flash --local-dir inclusionAI/Ming-3.0-audio-flash \u0026\u0026 export MODEL_PATH=\"$PWD/inclusionAI/Ming-3.0-audio-flash\" \u0026\u0026 pip install -r requirements-vllm-omni.txt \u0026\u0026 MING_OUTPUT_MODE=asr ./scripts/serve_vllm_omni.sh \"$MODEL_PATH\"", "runBuild": "Ming-3.0-audio-flash", "url": "/tools/ming-3-0-audio-flash", "builds": [ { "name": "Ming-3.0-audio-flash", "url": "https://huggingface.co/inclusionAI/Ming-3.0-audio-flash", "precision": "bf16", "format": "safetensors", "sizeGB": 259.3, "memoryGB": 512 } ] }, { "slug": "qwen3-omni-30b-a3b-instruct", "name": "Qwen3-Omni-30B-A3B-Instruct", "grade": "E", "score": 41.8, "tasks": [ "agents", "general", "vision" ], "params": "35.3B (3B active)", "context": 65536, "goodFor": "Omni-modal assistant that listens, watches and reads, then answers in text or speech, for voice agents, transcription, translation and audio-visual questions.", "licence": "Apache-2.0", "run": "vllm serve Qwen/Qwen3-Omni-30B-A3B-Instruct --dtype bfloat16 --max-model-len 32768 --allowed-local-media-path \"$(pwd)/media\"", "runBuild": "Qwen3-Omni-30B-A3B-Instruct", "url": "/tools/qwen3-omni-30b-a3b-instruct", "builds": [ { "name": "Qwen3-Omni-30B-A3B-Instruct", "url": "https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct", "precision": "bf16", "format": "safetensors", "sizeGB": 70.5, "memoryGB": 96 } ] }, { "slug": "granite-embedding-small-english-r2", "name": "Granite Embedding Small English R2", "grade": "E", "score": 41.7, "tasks": [ "embed" ], "paramsB": 0.047, "params": "47M", "context": 8192, "goodFor": "Fast English text and code retrieval embeddings on any hardware, with long inputs up to 8,192 tokens and a permissive data story.", "licence": "Apache-2.0", "run": "vllm serve ibm-granite/granite-embedding-small-english-r2 --runner pooling", "runBuild": "granite-embedding-small-english-r2", "url": "/tools/granite-embedding-small-english-r2", "builds": [ { "name": "granite-embedding-small-english-r2", "url": "https://huggingface.co/ibm-granite/granite-embedding-small-english-r2", "precision": "bf16", "format": "safetensors", "sizeGB": 0.1, "memoryGB": 8 } ] } ], "tasks": [ { "key": "agents", "label": "Agents and tool calling", "llm": true, "models": 77 }, { "key": "general", "label": "Chat and reasoning", "llm": true, "models": 86 }, { "key": "coding", "label": "Coding", "llm": true, "models": 17 }, { "key": "vision", "label": "Images and video in", "llm": true, "models": 49 }, { "key": "docs", "label": "Documents and OCR", "llm": false, "models": 3 }, { "key": "stt", "label": "Speech to text", "llm": false, "models": 10 }, { "key": "tts", "label": "Text to speech", "llm": false, "models": 8 }, { "key": "embed", "label": "Embeddings and search", "llm": false, "models": 9 }, { "key": "decide", "label": "Decisions with probabilities", "llm": true, "models": 5 } ], "worldUSDPerKWh": 0.179 }, "kind": "anchor.page", "links": { "api": "https://www.anchorterminal.com/api/v1/index.json", "html": "https://www.anchorterminal.com/open-weights/calculator/", "json": "https://www.anchorterminal.com/open-weights/calculator/index.json", "llms": "https://www.anchorterminal.com/llms.txt", "markdown": "https://www.anchorterminal.com/open-weights/calculator/index.md", "mcp": "https://www.anchorterminal.com/mcp", "slim": "https://www.anchorterminal.com/open-weights/calculator/index.min.md" }, "markdown": "Which open-weight model to run on your own machine, for the job you pick, and what it costs in power. We guess your hardware from what your browser says and your electricity price from your time zone, and you can change both. 120 graded models, each with the builds that fit.\n\nThe page runs the sum in the browser. This twin carries the same data and the method, so an agent can do it directly. The whole data set is in the page's JSON twin (https://www.anchorterminal.com/open-weights/calculator/index.json).\n\n## How it works\n\n- A model fits when the build's memory need, the file plus a fifth for the runtime and a working context, is no more than your memory times the share a model can have (92% of a graphics card, two thirds to three quarters of unified memory, 60% of a computer's RAM).\n- The pick is the best-graded model for the job with a build that fits and runs on your kind of machine, in its 8-bit build if that fits, then 6, 5 and 4-bit. A 16-bit build is picked only when nothing quantised fits, since it runs at about half the speed for little gain.\n- Speed is an estimate for one stream, from your memory bandwidth and the gigabytes read for each token (the build's size times the share of parameters active at once). Real speed depends on the runtime, the context length and the settings.\n- Cost is power under load times hours a day, at your electricity price, and per million tokens at the estimated speed. It leaves out the machine itself.\n\nNothing a browser reports is sent to us.\n\n## The machines it knows\n\nMemory and bandwidth are the makers' published figures. Power is a graphics card's board power, or our estimate for a whole machine under load (est.).\n\n| Machine | Memory | Usable share | Bandwidth | Power |\n| --- | --- | --- | --- | --- |\n| NVIDIA RTX 3060 12 GB | 12 GB | 92% | 360 GB/s | 170 W |\n| NVIDIA RTX 4060 Ti 16 GB | 16 GB | 92% | 288 GB/s | 165 W |\n| NVIDIA RTX 3090 24 GB | 24 GB | 92% | 936 GB/s | 350 W |\n| NVIDIA RTX 4090 24 GB | 24 GB | 92% | 1008 GB/s | 450 W |\n| NVIDIA RTX 5070 Ti 16 GB | 16 GB | 92% | 896 GB/s | 300 W |\n| NVIDIA RTX 5080 16 GB | 16 GB | 92% | 960 GB/s | 360 W |\n| NVIDIA RTX 5090 32 GB | 32 GB | 92% | 1792 GB/s | 575 W |\n| NVIDIA RTX PRO 6000 Blackwell 96 GB | 96 GB | 92% | 1792 GB/s | 600 W |\n| Apple M1, M2 or M3 | 8, 16, 24 GB | 67% | 100 GB/s | 30 W est. |\n| Apple M1 Pro, M2 Pro or M3 Pro | 16, 18, 32, 36 GB | 67% | 150 GB/s | 50 W est. |\n| Apple M1 Max, M2 Max or M3 Max | 32, 36, 48, 64, 96, 128 GB | 75% | 400 GB/s | 80 W est. |\n| Apple M1 Ultra or M2 Ultra | 64, 128, 192 GB | 75% | 800 GB/s | 160 W est. |\n| Apple M4 | 16, 24, 32 GB | 67% | 120 GB/s | 30 W est. |\n| Apple M4 Pro | 24, 48, 64 GB | 67% | 273 GB/s | 60 W est. |\n| Apple M4 Max | 36, 48, 64, 128 GB | 75% | 546 GB/s | 90 W est. |\n| Apple M3 Ultra | 96, 256, 512 GB | 75% | 819 GB/s | 180 W est. |\n| Apple M5 | 16, 24, 32 GB | 67% | 153 GB/s | 30 W est. |\n| Apple M5 Pro | 24, 48, 64 GB | 67% | 307 GB/s | 65 W est. |\n| Apple M5 Max | 36, 48, 64, 128 GB | 75% | 614 GB/s | 100 W est. |\n| AMD Ryzen AI Max+ 395 (Strix Halo) | 64, 96, 128 GB | 75% | 256 GB/s | 140 W est. |\n| NVIDIA DGX Spark | 128 GB | 85% | 273 GB/s | 200 W est. |\n| Any computer, on the CPU | 8, 16, 32, 64, 128 GB | 60% | 80 GB/s | 100 W est. |\n\n## Electricity prices\n\nHousehold prices in US dollars a kWh, GlobalPetrolPrices' 2023 to 2026 averages (https://www.globalpetrolprices.com/electricity_prices/), read on 9 November 2026. World average (Q3 2026): $0.179.\n\n| Country | $ a kWh |\n| --- | --- |\n| Argentina | 0.092 |\n| Australia | 0.259 |\n| Austria | 0.356 |\n| Belgium | 0.409 |\n| Brazil | 0.166 |\n| Canada | 0.123 |\n| China | 0.076 |\n| Czechia | 0.350 |\n| Denmark | 0.352 |\n| Finland | 0.177 |\n| France | 0.276 |\n| Germany | 0.409 |\n| Hong Kong | 0.184 |\n| India | 0.076 |\n| Indonesia | 0.090 |\n| Ireland | 0.453 |\n| Israel | 0.188 |\n| Italy | 0.415 |\n| Japan | 0.227 |\n| Mexico | 0.110 |\n| Netherlands | 0.287 |\n| New Zealand | 0.213 |\n| Nigeria | 0.036 |\n| Norway | 0.169 |\n| Philippines | 0.211 |\n| Poland | 0.236 |\n| Portugal | 0.240 |\n| Romania | 0.233 |\n| Singapore | 0.236 |\n| South Africa | 0.215 |\n| South Korea | 0.126 |\n| Spain | 0.259 |\n| Sweden | 0.247 |\n| Switzerland | 0.367 |\n| Taiwan | 0.099 |\n| Türkiye | 0.067 |\n| Ukraine | 0.085 |\n| United Arab Emirates | 0.080 |\n| United Kingdom | 0.405 |\n| United States | 0.190 |\n| Vietnam | 0.078 |\n\n## Models by job\n\nBest grade first, with the smallest memory any build needs. Each model's builds, with their memory needs, are on its listing and in this page's JSON twin.\n\n### Agents and tool calling\n\n| Model | Grade | Parameters | Smallest build needs |\n| --- | --- | --- | --- |\n| [Gemma 4 31B IT](https://www.anchorterminal.com/tools/gemma-4-31b-it.md) | BB 74.1 | 30.7B | 24 GB |\n| [Gemma 4 12B IT](https://www.anchorterminal.com/tools/gemma-4-12b-it.md) | BB 72.3 | 11.95B | 12 GB |\n| [Gemma 4 26B A4B IT](https://www.anchorterminal.com/tools/gemma-4-26b-a4b-it.md) | BB 72.2 | 25.2B | 24 GB |\n| [gpt-oss-120b](https://www.anchorterminal.com/tools/gpt-oss-120b.md) | BB 70.2 | 116.83B | 96 GB |\n| [Muse Glimmer 30B](https://www.anchorterminal.com/tools/muse-glimmer-30b.md) | BB 70 | 29.6B | 24 GB |\n| [gpt-oss-20b](https://www.anchorterminal.com/tools/gpt-oss-20b.md) | B 69.9 | 20.91B | 24 GB |\n| [Nemotron 3 Ultra 550B A55B](https://www.anchorterminal.com/tools/nemotron-3-ultra-550b-a55b.md) | B 69.3 | 550B | 512 GB |\n| [Qwen3.8-27B](https://www.anchorterminal.com/tools/qwen3-8-27b.md) | B 68.6 | 27B | 24 GB |\n| [Nemotron 3 Nano 30B A3B](https://www.anchorterminal.com/tools/nemotron-3-nano-30b-a3b.md) | B 67.6 | 30B | 24 GB |\n| [Inkling-Small](https://www.anchorterminal.com/tools/inkling-small.md) | B 67 | 276B | 256 GB |\n| [Nemotron 3 Super 120B A12B](https://www.anchorterminal.com/tools/nemotron-3-super-120b-a12b.md) | B 66.7 | 120B | 128 GB |\n| [Qwen3.6-27B](https://www.anchorterminal.com/tools/qwen3-6-27b.md) | B 66.7 | 27B | 24 GB |\n| [GLM-5.3-Flash](https://www.anchorterminal.com/tools/glm-5-3-flash.md) | B 65.1 | 320B | 256 GB |\n| [Inkling](https://www.anchorterminal.com/tools/inkling.md) | B 64.7 | 975B | 1024 GB |\n| [Nemotron 3.5 Lightning 30B A3B](https://www.anchorterminal.com/tools/nemotron-3-5-lightning-30b-a3b.md) | B 64.7 | 30B | 24 GB |\n| [GLM-5.3](https://www.anchorterminal.com/tools/glm-5-3.md) | B 64.6 | 744B | 1024 GB |\n| [Qwen3.5-397B-A17B](https://www.anchorterminal.com/tools/qwen3-5-397b-a17b.md) | B 64.1 | 397B total, 17B active | 512 GB |\n| [Gemma 4 E4B IT](https://www.anchorterminal.com/tools/gemma-4-e4b-it.md) | B 63.2 | 8B | 8 GB |\n| [Qwen3.5-122B-A10B](https://www.anchorterminal.com/tools/qwen3-5-122b-a10b.md) | B 63.1 | 122B total, 10B active | 96 GB |\n| [GLM-4.7-Flash](https://www.anchorterminal.com/tools/glm-4-7-flash.md) | B 63 | 30B | 24 GB |\n| [Qwen3.6-35B-A3B](https://www.anchorterminal.com/tools/qwen3-6-35b-a3b.md) | B 62.7 | 35B | 32 GB |\n| [Hy3](https://www.anchorterminal.com/tools/hy3.md) | B 62.6 | 295B | 128 GB |\n| [Kimi K3](https://www.anchorterminal.com/tools/kimi-k3.md) | B 62.4 | 2.8T | 2048 GB |\n| [DeepSeek-V4-Flash-0731](https://www.anchorterminal.com/tools/deepseek-v4-flash-0731.md) | C 61.9 | 284B | 256 GB |\n| [Qwen3.8-Flash-Next](https://www.anchorterminal.com/tools/qwen3-8-flash-next.md) | C 61.8 | 125B | 128 GB |\n| [DeepSeek V4.1 Flash](https://www.anchorterminal.com/tools/deepseek-v4-1-flash.md) | C 61.5 | 552B | 1024 GB |\n| [Nemotron 3 Nano Omni 30B A3B Reasoning](https://www.anchorterminal.com/tools/nemotron-3-nano-omni-30b-a3b-reasoning.md) | C 61.2 | 31B | 32 GB |\n| [Qwen3.5-4B](https://www.anchorterminal.com/tools/qwen3-5-4b.md) | C 61.2 | 4B | 8 GB |\n| [Agents-A1-4B](https://www.anchorterminal.com/tools/agents-a1-4b.md) | C 61 | 4.5B | 8 GB |\n| [Apertus v1.5 8B](https://www.anchorterminal.com/tools/apertus-v1-5-8b.md) | C 61 | 8.9B | 24 GB |\n| [Apodex-1.1-mini](https://www.anchorterminal.com/tools/apodex-1-1-mini.md) | C 61 | 35B (3B active) | 32 GB |\n| [Kolibri-1](https://www.anchorterminal.com/tools/kolibri-1.md) | C 61 | 78.1B | 96 GB |\n| [Laguna S 2.1](https://www.anchorterminal.com/tools/laguna-s-2-1.md) | C 61 | 117.6B total, 8B active | 96 GB |\n| [Laguna XS 2.1](https://www.anchorterminal.com/tools/laguna-xs-2-1.md) | C 61 | 33.4B total, 3B active | 32 GB |\n| [MiMo-V2.6 Distill Qwen 9B](https://www.anchorterminal.com/tools/mimo-v2-6-distill-qwen-9b.md) | C 61 | 9.41B | 8 GB |\n| [NVIDIA Nemotron 3.5 Bolt E4B 4GB](https://www.anchorterminal.com/tools/nemotron-3-5-bolt-e4b-4gb.md) | C 61 | 8B | 8 GB |\n| [Ornith-1.5-35B-A3B](https://www.anchorterminal.com/tools/ornith-1-5-35b-a3b.md) | C 61 | 35B (3B active) | 24 GB |\n| [Ornith-1.5-9B](https://www.anchorterminal.com/tools/ornith-1-5-9b.md) | C 61 | 9B | 8 GB |\n| [Qwen3-VL-2B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-2b-instruct.md) | C 61 | 2.13B | 8 GB |\n| [Qwopus3.8-27B-Flash-V2](https://www.anchorterminal.com/tools/qwopus3-8-27b-flash-v2.md) | C 61 | 27.8B | 24 GB |\n| [Recursive Flint](https://www.anchorterminal.com/tools/recursive-flint.md) | C 61 | 39.8B total, about 2.5B active | 32 GB |\n| [SmolLM3-3B](https://www.anchorterminal.com/tools/smollm3-3b.md) | C 61 | 3.08B | 8 GB |\n| [Spark-X2.6-4B](https://www.anchorterminal.com/tools/spark-x2-6-4b.md) | C 61 | 4.11B | 8 GB |\n| [Swift 1.5 Qwen3.8-27B](https://www.anchorterminal.com/tools/swift-1-5-qwen3-8-27b.md) | C 61 | 27B | 24 GB |\n| [ThinkingCap (Qwen3.8 27B)](https://www.anchorterminal.com/tools/thinkingcap-qwen3-8-27b.md) | C 61 | 27B | 24 GB |\n| [occamy-1.0](https://www.anchorterminal.com/tools/occamy-1-0.md) | C 61 | 35B (3B active) | 24 GB |\n| [Nanbeige4.3-33B](https://www.anchorterminal.com/tools/nanbeige-4-3-33b.md) | C 60.8 | 33B | 32 GB |\n| [Qwen3.5-9B](https://www.anchorterminal.com/tools/qwen3-5-9b.md) | C 60.5 | 9B | 8 GB |\n| [Gemma 4 E2B IT](https://www.anchorterminal.com/tools/gemma-4-e2b-it.md) | C 60.4 | 5.1B | 8 GB |\n| [GLM-4.5-Air](https://www.anchorterminal.com/tools/glm-4-5-air.md) | C 60.3 | 106B | 192 GB |\n| [Granite 4.2 3B](https://www.anchorterminal.com/tools/granite-4-2-3b.md) | C 59.8 | 3.66B | 8 GB |\n| [MiniCPM5-2B](https://www.anchorterminal.com/tools/minicpm5-2b.md) | C 59.7 | 2.5B | 8 GB |\n| [MiniMax-M2.5](https://www.anchorterminal.com/tools/minimax-m2-5.md) | C 59 | 228.7B | 512 GB |\n| [Ornith-1.5-397B](https://www.anchorterminal.com/tools/ornith-1-5-397b.md) | C 59 | 397B | 512 GB |\n| [DeepSeek-V3.2](https://www.anchorterminal.com/tools/deepseek-v3-2.md) | C 58.6 | 685B | 1024 GB |\n| [Qwen3-VL-30B-A3B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-30b-a3b-instruct.md) | C 58 | 31.1B (3B active) | 24 GB |\n| [MiniMax-M2.7](https://www.anchorterminal.com/tools/minimax-m2-7.md) | C 57.6 | 228.7B | 192 GB |\n| [Qwen3-Coder-Next](https://www.anchorterminal.com/tools/qwen3-coder-next.md) | C 57.6 | 80B total, 3B active | 64 GB |\n| [Nemotron 3 Nano 4B](https://www.anchorterminal.com/tools/nemotron-3-nano-4b.md) | C 56.8 | 4B | 8 GB |\n| [Laguna M.1](https://www.anchorterminal.com/tools/laguna-m-1.md) | C 56.7 | 225.8B total, 23B active | 192 GB |\n| [Qwen3-30B-A3B-Instruct-2507](https://www.anchorterminal.com/tools/qwen3-30b-a3b-instruct-2507.md) | C 56.6 | 30.5B (3.3B active) | 24 GB |\n| [Qwen3-VL-32B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-32b-instruct.md) | C 56.4 | 33.36B | 32 GB |\n| [BigBang-v1](https://www.anchorterminal.com/tools/bigbang-v1.md) | C 56.3 | 35B (3B active) | 96 GB |\n| [Ling-3.0-tiny](https://www.anchorterminal.com/tools/ling-3-0-tiny.md) | C 56.2 | 7.9B (1.3B active) | 8 GB |\n| [Qwen3-Coder-30B-A3B-Instruct](https://www.anchorterminal.com/tools/qwen3-coder-30b-a3b-instruct.md) | C 55.7 | 30.5B (3.3B active) | 24 GB |\n| [Qwen3-VL-8B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-8b-instruct.md) | C 54.9 | 8.77B | 8 GB |\n| [Mistral Large 4](https://www.anchorterminal.com/tools/mistral-large-4.md) | C 54.6 | ~1T total (1.05T per Mistral's docs), 52B active | 1024 GB |\n| [Qwen3-4B-Instruct-2507](https://www.anchorterminal.com/tools/qwen3-4b-instruct-2507.md) | C 54.5 | 4B | 8 GB |\n| [MiniMax-M3.1-Flash](https://www.anchorterminal.com/tools/minimax-m3-1-flash.md) | C 54.3 | 428B | 512 GB |\n| [Swift 1.5 Qwen3.8-Flash-Next](https://www.anchorterminal.com/tools/swift-1-5-qwen3-8-flash-next.md) | C 54.2 | 125B | 128 GB |\n| [Ripple-65B Preview](https://www.anchorterminal.com/tools/ripple-65b-preview.md) | D 53.3 | 65B | 69 GB |\n| [Qwen3-VL-4B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-4b-instruct.md) | D 52.6 | 4.44B | 8 GB |\n| [DiffusionGemma 26B A4B IT](https://www.anchorterminal.com/tools/diffusiongemma-26b-a4b-it.md) | D 51.3 | 25.2B | 64 GB |\n| [Qwen3.5-2B](https://www.anchorterminal.com/tools/qwen3-5-2b.md) | D 50.5 | 2B | 8 GB |\n| [LFM2.5-2.6B](https://www.anchorterminal.com/tools/lfm2-5-2-6b.md) | D 50.3 | 2.69B | 8 GB |\n| [Qwen3.5-0.8B](https://www.anchorterminal.com/tools/qwen3-5-0-8b.md) | D 46.7 | 0.8B | 8 GB |\n| [Qwen3-Omni-30B-A3B-Instruct](https://www.anchorterminal.com/tools/qwen3-omni-30b-a3b-instruct.md) | E 41.8 | 35.3B (3B active) | 96 GB |\n\n### Chat and reasoning\n\n| Model | Grade | Parameters | Smallest build needs |\n| --- | --- | --- | --- |\n| [Gemma 4 31B IT](https://www.anchorterminal.com/tools/gemma-4-31b-it.md) | BB 74.1 | 30.7B | 24 GB |\n| [Gemma 4 12B IT](https://www.anchorterminal.com/tools/gemma-4-12b-it.md) | BB 72.3 | 11.95B | 12 GB |\n| [Gemma 4 26B A4B IT](https://www.anchorterminal.com/tools/gemma-4-26b-a4b-it.md) | BB 72.2 | 25.2B | 24 GB |\n| [gpt-oss-120b](https://www.anchorterminal.com/tools/gpt-oss-120b.md) | BB 70.2 | 116.83B | 96 GB |\n| [Muse Glimmer 30B](https://www.anchorterminal.com/tools/muse-glimmer-30b.md) | BB 70 | 29.6B | 24 GB |\n| [gpt-oss-20b](https://www.anchorterminal.com/tools/gpt-oss-20b.md) | B 69.9 | 20.91B | 24 GB |\n| [Nemotron 3 Ultra 550B A55B](https://www.anchorterminal.com/tools/nemotron-3-ultra-550b-a55b.md) | B 69.3 | 550B | 512 GB |\n| [Qwen3.8-27B](https://www.anchorterminal.com/tools/qwen3-8-27b.md) | B 68.6 | 27B | 24 GB |\n| [Nemotron 3 Nano 30B A3B](https://www.anchorterminal.com/tools/nemotron-3-nano-30b-a3b.md) | B 67.6 | 30B | 24 GB |\n| [Inkling-Small](https://www.anchorterminal.com/tools/inkling-small.md) | B 67 | 276B | 256 GB |\n| [Nemotron 3 Super 120B A12B](https://www.anchorterminal.com/tools/nemotron-3-super-120b-a12b.md) | B 66.7 | 120B | 128 GB |\n| [Qwen3.6-27B](https://www.anchorterminal.com/tools/qwen3-6-27b.md) | B 66.7 | 27B | 24 GB |\n| [GLM-5.3-Flash](https://www.anchorterminal.com/tools/glm-5-3-flash.md) | B 65.1 | 320B | 256 GB |\n| [Inkling](https://www.anchorterminal.com/tools/inkling.md) | B 64.7 | 975B | 1024 GB |\n| [Nemotron 3.5 Lightning 30B A3B](https://www.anchorterminal.com/tools/nemotron-3-5-lightning-30b-a3b.md) | B 64.7 | 30B | 24 GB |\n| [GLM-5.3](https://www.anchorterminal.com/tools/glm-5-3.md) | B 64.6 | 744B | 1024 GB |\n| [Qwen3.5-397B-A17B](https://www.anchorterminal.com/tools/qwen3-5-397b-a17b.md) | B 64.1 | 397B total, 17B active | 512 GB |\n| [Gemma 4 E4B IT](https://www.anchorterminal.com/tools/gemma-4-e4b-it.md) | B 63.2 | 8B | 8 GB |\n| [Qwen3.5-122B-A10B](https://www.anchorterminal.com/tools/qwen3-5-122b-a10b.md) | B 63.1 | 122B total, 10B active | 96 GB |\n| [GLM-4.7-Flash](https://www.anchorterminal.com/tools/glm-4-7-flash.md) | B 63 | 30B | 24 GB |\n| [Qwen3.6-35B-A3B](https://www.anchorterminal.com/tools/qwen3-6-35b-a3b.md) | B 62.7 | 35B | 32 GB |\n| [Hy3](https://www.anchorterminal.com/tools/hy3.md) | B 62.6 | 295B | 128 GB |\n| [Kimi K3](https://www.anchorterminal.com/tools/kimi-k3.md) | B 62.4 | 2.8T | 2048 GB |\n| [DeepSeek-V4-Flash-0731](https://www.anchorterminal.com/tools/deepseek-v4-flash-0731.md) | C 61.9 | 284B | 256 GB |\n| [Qwen3.8-Flash-Next](https://www.anchorterminal.com/tools/qwen3-8-flash-next.md) | C 61.8 | 125B | 128 GB |\n| [DeepSeek V4.1 Flash](https://www.anchorterminal.com/tools/deepseek-v4-1-flash.md) | C 61.5 | 552B | 1024 GB |\n| [Nemotron 3 Nano Omni 30B A3B Reasoning](https://www.anchorterminal.com/tools/nemotron-3-nano-omni-30b-a3b-reasoning.md) | C 61.2 | 31B | 32 GB |\n| [Qwen3.5-4B](https://www.anchorterminal.com/tools/qwen3-5-4b.md) | C 61.2 | 4B | 8 GB |\n| [Agents-A1-4B](https://www.anchorterminal.com/tools/agents-a1-4b.md) | C 61 | 4.5B | 8 GB |\n| [Apertus v1.5 8B](https://www.anchorterminal.com/tools/apertus-v1-5-8b.md) | C 61 | 8.9B | 24 GB |\n| [Apodex-1.1-mini](https://www.anchorterminal.com/tools/apodex-1-1-mini.md) | C 61 | 35B (3B active) | 32 GB |\n| [Carbon-X](https://www.anchorterminal.com/tools/carbon-x.md) | C 61 | 14.8B | 12 GB |\n| [Darwin-36B-Opus](https://www.anchorterminal.com/tools/darwin-36b-opus.md) | C 61 | 35B (3B active) | 12 GB |\n| [Kolibri-1](https://www.anchorterminal.com/tools/kolibri-1.md) | C 61 | 78.1B | 96 GB |\n| [Laguna S 2.1](https://www.anchorterminal.com/tools/laguna-s-2-1.md) | C 61 | 117.6B total, 8B active | 96 GB |\n| [Laguna XS 2.1](https://www.anchorterminal.com/tools/laguna-xs-2-1.md) | C 61 | 33.4B total, 3B active | 32 GB |\n| [MiMo-V2.6 Distill Qwen 9B](https://www.anchorterminal.com/tools/mimo-v2-6-distill-qwen-9b.md) | C 61 | 9.41B | 8 GB |\n| [MiniCPM-o 4.5](https://www.anchorterminal.com/tools/minicpm-o-4-5.md) | C 61 | 9.4B | 8 GB |\n| [NVIDIA Nemotron 3.5 Bolt E4B 4GB](https://www.anchorterminal.com/tools/nemotron-3-5-bolt-e4b-4gb.md) | C 61 | 8B | 8 GB |\n| [Ornith-1.5-35B-A3B](https://www.anchorterminal.com/tools/ornith-1-5-35b-a3b.md) | C 61 | 35B (3B active) | 24 GB |\n| [Ornith-1.5-9B](https://www.anchorterminal.com/tools/ornith-1-5-9b.md) | C 61 | 9B | 8 GB |\n| [Qwen3-VL-2B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-2b-instruct.md) | C 61 | 2.13B | 8 GB |\n| [Qwopus3.8-27B-Flash-V2](https://www.anchorterminal.com/tools/qwopus3-8-27b-flash-v2.md) | C 61 | 27.8B | 24 GB |\n| [Recursive Flint](https://www.anchorterminal.com/tools/recursive-flint.md) | C 61 | 39.8B total, about 2.5B active | 32 GB |\n| [SmolLM3-3B](https://www.anchorterminal.com/tools/smollm3-3b.md) | C 61 | 3.08B | 8 GB |\n| [Spark-X2.6-4B](https://www.anchorterminal.com/tools/spark-x2-6-4b.md) | C 61 | 4.11B | 8 GB |\n| [Swift 1.5 Qwen3.8-27B](https://www.anchorterminal.com/tools/swift-1-5-qwen3-8-27b.md) | C 61 | 27B | 24 GB |\n| [ThinkingCap (Qwen3.8 27B)](https://www.anchorterminal.com/tools/thinkingcap-qwen3-8-27b.md) | C 61 | 27B | 24 GB |\n| [occamy-1.0](https://www.anchorterminal.com/tools/occamy-1-0.md) | C 61 | 35B (3B active) | 24 GB |\n| [Nanbeige4.3-33B](https://www.anchorterminal.com/tools/nanbeige-4-3-33b.md) | C 60.8 | 33B | 32 GB |\n| [Qwen3.5-9B](https://www.anchorterminal.com/tools/qwen3-5-9b.md) | C 60.5 | 9B | 8 GB |\n| [Gemma 4 E2B IT](https://www.anchorterminal.com/tools/gemma-4-e2b-it.md) | C 60.4 | 5.1B | 8 GB |\n| [GLM-4.5-Air](https://www.anchorterminal.com/tools/glm-4-5-air.md) | C 60.3 | 106B | 192 GB |\n| [Granite 4.2 3B](https://www.anchorterminal.com/tools/granite-4-2-3b.md) | C 59.8 | 3.66B | 8 GB |\n| [MiniCPM5-2B](https://www.anchorterminal.com/tools/minicpm5-2b.md) | C 59.7 | 2.5B | 8 GB |\n| [Cosmos-Reason2-2B](https://www.anchorterminal.com/tools/cosmos-reason2-2b.md) | C 59.5 | 2.44B | 8 GB |\n| [MiniMax-M2.5](https://www.anchorterminal.com/tools/minimax-m2-5.md) | C 59 | 228.7B | 512 GB |\n| [Ornith-1.5-397B](https://www.anchorterminal.com/tools/ornith-1-5-397b.md) | C 59 | 397B | 512 GB |\n| [DeepSeek-V3.2](https://www.anchorterminal.com/tools/deepseek-v3-2.md) | C 58.6 | 685B | 1024 GB |\n| [Qwen3-VL-30B-A3B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-30b-a3b-instruct.md) | C 58 | 31.1B (3B active) | 24 GB |\n| [MiniMax-M2.7](https://www.anchorterminal.com/tools/minimax-m2-7.md) | C 57.6 | 228.7B | 192 GB |\n| [Qwen3-Coder-Next](https://www.anchorterminal.com/tools/qwen3-coder-next.md) | C 57.6 | 80B total, 3B active | 64 GB |\n| [Nemotron 3 Nano 4B](https://www.anchorterminal.com/tools/nemotron-3-nano-4b.md) | C 56.8 | 4B | 8 GB |\n| [Laguna M.1](https://www.anchorterminal.com/tools/laguna-m-1.md) | C 56.7 | 225.8B total, 23B active | 192 GB |\n| [Qwen3-30B-A3B-Instruct-2507](https://www.anchorterminal.com/tools/qwen3-30b-a3b-instruct-2507.md) | C 56.6 | 30.5B (3.3B active) | 24 GB |\n| [Qwen3-VL-32B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-32b-instruct.md) | C 56.4 | 33.36B | 32 GB |\n| [BigBang-v1](https://www.anchorterminal.com/tools/bigbang-v1.md) | C 56.3 | 35B (3B active) | 96 GB |\n| [Ling-3.0-tiny](https://www.anchorterminal.com/tools/ling-3-0-tiny.md) | C 56.2 | 7.9B (1.3B active) | 8 GB |\n| [LBS-27B](https://www.anchorterminal.com/tools/lbs-27b.md) | C 55.8 | 27B | 96 GB |\n| [Qwen3-Coder-30B-A3B-Instruct](https://www.anchorterminal.com/tools/qwen3-coder-30b-a3b-instruct.md) | C 55.7 | 30.5B (3.3B active) | 24 GB |\n| [Kronos](https://www.anchorterminal.com/tools/kronos.md) | C 55.2 | 4.1M to 102.3M | 8 GB |\n| [Qwen3-VL-8B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-8b-instruct.md) | C 54.9 | 8.77B | 8 GB |\n| [Mistral Large 4](https://www.anchorterminal.com/tools/mistral-large-4.md) | C 54.6 | ~1T total (1.05T per Mistral's docs), 52B active | 1024 GB |\n| [Qwen3-4B-Instruct-2507](https://www.anchorterminal.com/tools/qwen3-4b-instruct-2507.md) | C 54.5 | 4B | 8 GB |\n| [MiniMax-M3.1-Flash](https://www.anchorterminal.com/tools/minimax-m3-1-flash.md) | C 54.3 | 428B | 512 GB |\n| [Baidu Unlimited-OCR-2](https://www.anchorterminal.com/tools/unlimited-ocr-2.md) | C 54.2 | 3.3B (3B-A0.5B mixture-of-experts decoder per the report) | 12 GB |\n| [Swift 1.5 Qwen3.8-Flash-Next](https://www.anchorterminal.com/tools/swift-1-5-qwen3-8-flash-next.md) | C 54.2 | 125B | 128 GB |\n| [Ripple-65B Preview](https://www.anchorterminal.com/tools/ripple-65b-preview.md) | D 53.3 | 65B | 69 GB |\n| [InternVL3.5-GPT-OSS-20B-A4B-Preview](https://www.anchorterminal.com/tools/internvl3-5-gpt-oss-20b-a4b-preview.md) | D 53.2 | 21.2B | 64 GB |\n| [Qwen3-VL-4B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-4b-instruct.md) | D 52.6 | 4.44B | 8 GB |\n| [Olmo 3 7B Think](https://www.anchorterminal.com/tools/olmo-3-7b-think.md) | D 52.3 | 7.3B | 8 GB |\n| [DiffusionGemma 26B A4B IT](https://www.anchorterminal.com/tools/diffusiongemma-26b-a4b-it.md) | D 51.3 | 25.2B | 64 GB |\n| [Qwen3.5-2B](https://www.anchorterminal.com/tools/qwen3-5-2b.md) | D 50.5 | 2B | 8 GB |\n| [LFM2.5-2.6B](https://www.anchorterminal.com/tools/lfm2-5-2-6b.md) | D 50.3 | 2.69B | 8 GB |\n| [Qwen3.5-0.8B](https://www.anchorterminal.com/tools/qwen3-5-0-8b.md) | D 46.7 | 0.8B | 8 GB |\n| [Qwen3-Omni-30B-A3B-Instruct](https://www.anchorterminal.com/tools/qwen3-omni-30b-a3b-instruct.md) | E 41.8 | 35.3B (3B active) | 96 GB |\n\n### Coding\n\n| Model | Grade | Parameters | Smallest build needs |\n| --- | --- | --- | --- |\n| [Qwen3.6-27B](https://www.anchorterminal.com/tools/qwen3-6-27b.md) | B 66.7 | 27B | 24 GB |\n| [GLM-5.3](https://www.anchorterminal.com/tools/glm-5-3.md) | B 64.6 | 744B | 1024 GB |\n| [Laguna S 2.1](https://www.anchorterminal.com/tools/laguna-s-2-1.md) | C 61 | 117.6B total, 8B active | 96 GB |\n| [Laguna XS 2.1](https://www.anchorterminal.com/tools/laguna-xs-2-1.md) | C 61 | 33.4B total, 3B active | 32 GB |\n| [Ornith-1.5-35B-A3B](https://www.anchorterminal.com/tools/ornith-1-5-35b-a3b.md) | C 61 | 35B (3B active) | 24 GB |\n| [Ornith-1.5-9B](https://www.anchorterminal.com/tools/ornith-1-5-9b.md) | C 61 | 9B | 8 GB |\n| [Recursive Flint](https://www.anchorterminal.com/tools/recursive-flint.md) | C 61 | 39.8B total, about 2.5B active | 32 GB |\n| [occamy-1.0](https://www.anchorterminal.com/tools/occamy-1-0.md) | C 61 | 35B (3B active) | 24 GB |\n| [Nanbeige4.3-33B](https://www.anchorterminal.com/tools/nanbeige-4-3-33b.md) | C 60.8 | 33B | 32 GB |\n| [MiniMax-M2.5](https://www.anchorterminal.com/tools/minimax-m2-5.md) | C 59 | 228.7B | 512 GB |\n| [Ornith-1.5-397B](https://www.anchorterminal.com/tools/ornith-1-5-397b.md) | C 59 | 397B | 512 GB |\n| [MiniMax-M2.7](https://www.anchorterminal.com/tools/minimax-m2-7.md) | C 57.6 | 228.7B | 192 GB |\n| [Qwen3-Coder-Next](https://www.anchorterminal.com/tools/qwen3-coder-next.md) | C 57.6 | 80B total, 3B active | 64 GB |\n| [Laguna M.1](https://www.anchorterminal.com/tools/laguna-m-1.md) | C 56.7 | 225.8B total, 23B active | 192 GB |\n| [Qwen3-Coder-30B-A3B-Instruct](https://www.anchorterminal.com/tools/qwen3-coder-30b-a3b-instruct.md) | C 55.7 | 30.5B (3.3B active) | 24 GB |\n| [MiniMax-M3.1-Flash](https://www.anchorterminal.com/tools/minimax-m3-1-flash.md) | C 54.3 | 428B | 512 GB |\n| [Ripple-65B Preview](https://www.anchorterminal.com/tools/ripple-65b-preview.md) | D 53.3 | 65B | 69 GB |\n\n### Images and video in\n\n| Model | Grade | Parameters | Smallest build needs |\n| --- | --- | --- | --- |\n| [Gemma 4 31B IT](https://www.anchorterminal.com/tools/gemma-4-31b-it.md) | BB 74.1 | 30.7B | 24 GB |\n| [Gemma 4 12B IT](https://www.anchorterminal.com/tools/gemma-4-12b-it.md) | BB 72.3 | 11.95B | 12 GB |\n| [Gemma 4 26B A4B IT](https://www.anchorterminal.com/tools/gemma-4-26b-a4b-it.md) | BB 72.2 | 25.2B | 24 GB |\n| [Muse Glimmer 30B](https://www.anchorterminal.com/tools/muse-glimmer-30b.md) | BB 70 | 29.6B | 24 GB |\n| [Qwen3.8-27B](https://www.anchorterminal.com/tools/qwen3-8-27b.md) | B 68.6 | 27B | 24 GB |\n| [Inkling-Small](https://www.anchorterminal.com/tools/inkling-small.md) | B 67 | 276B | 256 GB |\n| [Qwen3.6-27B](https://www.anchorterminal.com/tools/qwen3-6-27b.md) | B 66.7 | 27B | 24 GB |\n| [GLM-5.3-Flash](https://www.anchorterminal.com/tools/glm-5-3-flash.md) | B 65.1 | 320B | 256 GB |\n| [Inkling](https://www.anchorterminal.com/tools/inkling.md) | B 64.7 | 975B | 1024 GB |\n| [Qwen3.5-397B-A17B](https://www.anchorterminal.com/tools/qwen3-5-397b-a17b.md) | B 64.1 | 397B total, 17B active | 512 GB |\n| [Gemma 4 E4B IT](https://www.anchorterminal.com/tools/gemma-4-e4b-it.md) | B 63.2 | 8B | 8 GB |\n| [Qwen3.5-122B-A10B](https://www.anchorterminal.com/tools/qwen3-5-122b-a10b.md) | B 63.1 | 122B total, 10B active | 96 GB |\n| [Qwen3.6-35B-A3B](https://www.anchorterminal.com/tools/qwen3-6-35b-a3b.md) | B 62.7 | 35B | 32 GB |\n| [Kimi K3](https://www.anchorterminal.com/tools/kimi-k3.md) | B 62.4 | 2.8T | 2048 GB |\n| [Qwen3.8-Flash-Next](https://www.anchorterminal.com/tools/qwen3-8-flash-next.md) | C 61.8 | 125B | 128 GB |\n| [DeepSeek V4.1 Flash](https://www.anchorterminal.com/tools/deepseek-v4-1-flash.md) | C 61.5 | 552B | 1024 GB |\n| [Nemotron 3 Nano Omni 30B A3B Reasoning](https://www.anchorterminal.com/tools/nemotron-3-nano-omni-30b-a3b-reasoning.md) | C 61.2 | 31B | 32 GB |\n| [Qwen3.5-4B](https://www.anchorterminal.com/tools/qwen3-5-4b.md) | C 61.2 | 4B | 8 GB |\n| [Agents-A1-4B](https://www.anchorterminal.com/tools/agents-a1-4b.md) | C 61 | 4.5B | 8 GB |\n| [Apertus v1.5 8B](https://www.anchorterminal.com/tools/apertus-v1-5-8b.md) | C 61 | 8.9B | 24 GB |\n| [Apodex-1.1-mini](https://www.anchorterminal.com/tools/apodex-1-1-mini.md) | C 61 | 35B (3B active) | 32 GB |\n| [MiMo-V2.6 Distill Qwen 9B](https://www.anchorterminal.com/tools/mimo-v2-6-distill-qwen-9b.md) | C 61 | 9.41B | 8 GB |\n| [MiniCPM-o 4.5](https://www.anchorterminal.com/tools/minicpm-o-4-5.md) | C 61 | 9.4B | 8 GB |\n| [Ornith-1.5-35B-A3B](https://www.anchorterminal.com/tools/ornith-1-5-35b-a3b.md) | C 61 | 35B (3B active) | 24 GB |\n| [Ornith-1.5-9B](https://www.anchorterminal.com/tools/ornith-1-5-9b.md) | C 61 | 9B | 8 GB |\n| [Qwen3-VL-2B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-2b-instruct.md) | C 61 | 2.13B | 8 GB |\n| [Qwopus3.8-27B-Flash-V2](https://www.anchorterminal.com/tools/qwopus3-8-27b-flash-v2.md) | C 61 | 27.8B | 24 GB |\n| [Swift 1.5 Qwen3.8-27B](https://www.anchorterminal.com/tools/swift-1-5-qwen3-8-27b.md) | C 61 | 27B | 24 GB |\n| [ThinkingCap (Qwen3.8 27B)](https://www.anchorterminal.com/tools/thinkingcap-qwen3-8-27b.md) | C 61 | 27B | 24 GB |\n| [occamy-1.0](https://www.anchorterminal.com/tools/occamy-1-0.md) | C 61 | 35B (3B active) | 24 GB |\n| [Qwen3.5-9B](https://www.anchorterminal.com/tools/qwen3-5-9b.md) | C 60.5 | 9B | 8 GB |\n| [Gemma 4 E2B IT](https://www.anchorterminal.com/tools/gemma-4-e2b-it.md) | C 60.4 | 5.1B | 8 GB |\n| [Cosmos-Reason2-2B](https://www.anchorterminal.com/tools/cosmos-reason2-2b.md) | C 59.5 | 2.44B | 8 GB |\n| [Ornith-1.5-397B](https://www.anchorterminal.com/tools/ornith-1-5-397b.md) | C 59 | 397B | 512 GB |\n| [Qwen3-VL-30B-A3B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-30b-a3b-instruct.md) | C 58 | 31.1B (3B active) | 24 GB |\n| [Qwen3-VL-32B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-32b-instruct.md) | C 56.4 | 33.36B | 32 GB |\n| [BigBang-v1](https://www.anchorterminal.com/tools/bigbang-v1.md) | C 56.3 | 35B (3B active) | 96 GB |\n| [LBS-27B](https://www.anchorterminal.com/tools/lbs-27b.md) | C 55.8 | 27B | 96 GB |\n| [Qwen3-VL-8B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-8b-instruct.md) | C 54.9 | 8.77B | 8 GB |\n| [Mistral Large 4](https://www.anchorterminal.com/tools/mistral-large-4.md) | C 54.6 | ~1T total (1.05T per Mistral's docs), 52B active | 1024 GB |\n| [MiniMax-M3.1-Flash](https://www.anchorterminal.com/tools/minimax-m3-1-flash.md) | C 54.3 | 428B | 512 GB |\n| [Baidu Unlimited-OCR-2](https://www.anchorterminal.com/tools/unlimited-ocr-2.md) | C 54.2 | 3.3B (3B-A0.5B mixture-of-experts decoder per the report) | 12 GB |\n| [Swift 1.5 Qwen3.8-Flash-Next](https://www.anchorterminal.com/tools/swift-1-5-qwen3-8-flash-next.md) | C 54.2 | 125B | 128 GB |\n| [InternVL3.5-GPT-OSS-20B-A4B-Preview](https://www.anchorterminal.com/tools/internvl3-5-gpt-oss-20b-a4b-preview.md) | D 53.2 | 21.2B | 64 GB |\n| [Qwen3-VL-4B-Instruct](https://www.anchorterminal.com/tools/qwen3-vl-4b-instruct.md) | D 52.6 | 4.44B | 8 GB |\n| [DiffusionGemma 26B A4B IT](https://www.anchorterminal.com/tools/diffusiongemma-26b-a4b-it.md) | D 51.3 | 25.2B | 64 GB |\n| [Qwen3.5-2B](https://www.anchorterminal.com/tools/qwen3-5-2b.md) | D 50.5 | 2B | 8 GB |\n| [Qwen3.5-0.8B](https://www.anchorterminal.com/tools/qwen3-5-0-8b.md) | D 46.7 | 0.8B | 8 GB |\n| [Qwen3-Omni-30B-A3B-Instruct](https://www.anchorterminal.com/tools/qwen3-omni-30b-a3b-instruct.md) | E 41.8 | 35.3B (3B active) | 96 GB |\n\n### Documents and OCR\n\n| Model | Grade | Parameters | Smallest build needs |\n| --- | --- | --- | --- |\n| [Nemotron 3.5 Parse](https://www.anchorterminal.com/tools/nemotron-3-5-parse.md) | B 69 | 0.63B | 8 GB |\n| [Xiaomi-OCR-0](https://www.anchorterminal.com/tools/xiaomi-ocr-0.md) | C 59 | 0.87B (873,438,784 by the Hub's count; 0.8B on the card) | 8 GB |\n| [Baidu Unlimited-OCR-2](https://www.anchorterminal.com/tools/unlimited-ocr-2.md) | C 54.2 | 3.3B (3B-A0.5B mixture-of-experts decoder per the report) | 12 GB |\n\n### Speech to text\n\n| Model | Grade | Parameters | Smallest build needs |\n| --- | --- | --- | --- |\n| [Parakeet TDT 0.6B v3](https://www.anchorterminal.com/tools/parakeet-tdt-0-6b-v3.md) | B 66.9 | 0.6B | 8 GB |\n| [Qwen3-ASR-1.7B](https://www.anchorterminal.com/tools/qwen3-asr-1-7b.md) | B 64.4 | 2.04B | 8 GB |\n| [Voxtral Mini 4B Realtime 2602](https://www.anchorterminal.com/tools/voxtral-mini-4b-realtime-2602.md) | B 62 | 4.43B | 8 GB |\n| [Qwen3-ASR-0.6B](https://www.anchorterminal.com/tools/qwen3-asr-0-6b.md) | C 61.2 | 0.78B | 8 GB |\n| [Cohere Transcribe 03-2026](https://www.anchorterminal.com/tools/cohere-transcribe-03-2026.md) | C 60.3 | 2.07B | 8 GB |\n| [Nemotron 3.5 ASR Streaming 0.6B](https://www.anchorterminal.com/tools/nemotron-3-5-asr-streaming-0-6b.md) | C 60.3 | 0.64B | 8 GB |\n| [Ming-3.0-asr-tiny](https://www.anchorterminal.com/tools/ming-3-0-asr-tiny.md) | C 54.8 | 8.87B | 24 GB |\n| [Ming-3.0-audio-tiny](https://www.anchorterminal.com/tools/ming-3-0-audio-tiny.md) | D 52.1 | 9.3B | 32 GB |\n| [Underdog Bark 3.0](https://www.anchorterminal.com/tools/underdog-bark-3.md) | D 48.2 | 0.6B | 8 GB |\n| [Ming-3.0-audio-flash](https://www.anchorterminal.com/tools/ming-3-0-audio-flash.md) | E 42.5 | 126B | 512 GB |\n\n### Text to speech\n\n| Model | Grade | Parameters | Smallest build needs |\n| --- | --- | --- | --- |\n| [Ming-3.0-tts-0.5B](https://www.anchorterminal.com/tools/ming-3-0-tts-0-5b.md) | C 59.3 | 0.5B | 8 GB |\n| [Qwen3-TTS 12Hz 1.7B CustomVoice](https://www.anchorterminal.com/tools/qwen3-tts-1-7b-customvoice.md) | C 59.2 | 1.92B (main checkpoint) plus the 12 Hz speech tokenizer | 8 GB |\n| [VieNeu-TTS v3 Turbo](https://www.anchorterminal.com/tools/vieneu-tts-v3-turbo.md) | C 58.5 | 0.13B | 8 GB |\n| [OmniVoice](https://www.anchorterminal.com/tools/omnivoice.md) | C 56.9 | 0.61B (main checkpoint) plus the Higgs Audio v2 tokenizer | 8 GB |\n| [Fish Audio Drama 3](https://www.anchorterminal.com/tools/fish-audio-drama-3.md) | C 55.6 | 35.9B (35B backbone with about 3B active per token, per the card; 35,899,643,008 by the Hub's count) | 32 GB |\n| [Ming-3.0-tts-tiny](https://www.anchorterminal.com/tools/ming-3-0-tts-tiny.md) | C 54.5 | 7.9B | 24 GB |\n| [Fish Audio S2 Pro](https://www.anchorterminal.com/tools/fish-audio-s2-pro.md) | D 52.5 | 4.56B (slow AR 4B plus fast AR 0.4B, BF16) plus the codec | 16 GB |\n| [Audio8-TTS-Realistic 0.6B](https://www.anchorterminal.com/tools/audio8-tts-realistic-0-6b.md) | D 51.5 | 0.6B (601,159,424 BF16 parameters by the Hub's count, without the bundled codec) | 8 GB |\n\n### Embeddings and search\n\n| Model | Grade | Parameters | Smallest build needs |\n| --- | --- | --- | --- |\n| [Qwen3-Embedding-8B](https://www.anchorterminal.com/tools/qwen3-embedding-8b.md) | BB 75.4 | 7.57B | 8 GB |\n| [Qwen3-Embedding-4B](https://www.anchorterminal.com/tools/qwen3-embedding-4b.md) | BB 73.6 | 4.02B | 8 GB |\n| [EmbeddingGemma 2](https://www.anchorterminal.com/tools/embeddinggemma-2.md) | B 66.4 | 740M | 8 GB |\n| [Qwen3-Embedding-0.6B](https://www.anchorterminal.com/tools/qwen3-embedding-0-6b.md) | B 62.5 | 0.6B | 8 GB |\n| [Qwen3-VL-Embedding-8B](https://www.anchorterminal.com/tools/qwen3-vl-embedding-8b.md) | B 62.1 | 8.14B | 24 GB |\n| [Qwen3-VL-Embedding-2B](https://www.anchorterminal.com/tools/qwen3-vl-embedding-2b.md) | C 57.7 | 2.13B | 8 GB |\n| [EmbeddingGemma 300M](https://www.anchorterminal.com/tools/embeddinggemma-300m.md) | D 48.9 | 308M | 8 GB |\n| [jina-embeddings-v5-omni-small](https://www.anchorterminal.com/tools/jina-embeddings-v5-omni-small.md) | D 48.8 | 1.63B | 8 GB |\n| [Granite Embedding Small English R2](https://www.anchorterminal.com/tools/granite-embedding-small-english-r2.md) | E 41.7 | 47M | 8 GB |\n\n### Decisions with probabilities\n\n| Model | Grade | Parameters | Smallest build needs |\n| --- | --- | --- | --- |\n| [JEV-27B-VL](https://www.anchorterminal.com/tools/jev-27b-vl.md) | C 59.5 | 27B | 96 GB |\n| [joey-1](https://www.anchorterminal.com/tools/joey-1.md) | C 57.1 | 2B | 8 GB |\n| [Drex Max v1](https://www.anchorterminal.com/tools/drex-max-v1.md) | C 55.5 | 27B | 24 GB |\n| [elephant-1](https://www.anchorterminal.com/tools/elephant-1.md) | D 53.6 | 9B | 24 GB |\n| [mammoth-1](https://www.anchorterminal.com/tools/mammoth-1.md) | D 48.6 | 27B | 96 GB |\n\n", "meta": { "attribution": "Anchor Terminal (https://www.anchorterminal.com)", "attributionRequired": true, "credit": "CC BY 4.0, credit Anchor Terminal and link the page or API address you took it from (https://www.anchorterminal.com/terms/#data)", "docs": "https://www.anchorterminal.com/docs/", "generatedAt": "2026-11-09", "license": "CC-BY-4.0", "method": "https://www.anchorterminal.com/benchmark/", "methodology": "0.7", "openapi": "https://www.anchorterminal.com/openapi.json", "preview": false, "run": "2026-10-01", "runLabel": "October 2026 research run", "terms": "https://www.anchorterminal.com/terms/#data" }, "page": { "breadcrumbs": [ { "name": "Home", "url": "https://www.anchorterminal.com/" }, { "name": "Open-weight models", "url": "https://www.anchorterminal.com/open-weights/" }, { "name": "Calculator", "url": "" } ], "description": "Which open-weight model to run on your own machine for the job you pick, with the build that fits, an estimate of its speed and what it costs in electricity, with your hardware and power price guessed for you.", "facts": [ "120 models", "22 machines", "power and cost" ], "h1": "What can I run locally?", "image": "https://www.anchorterminal.com/assets/og/default.png", "path": "/open-weights/calculator/", "published": "", "section": "tools", "title": "What can I run locally? Open-weight model calculator | Anchor Terminal", "toc": null, "updated": "2026-11-09", "url": "https://www.anchorterminal.com/open-weights/calculator/" }, "tokens": { "markdown": 8900, "slim": 8830 }, "version": 1 }