{ "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": "2026-11-11", "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" }, "tool": { "slug": "nemotron-3-5-transcribe-30b-a4b", "name": "Nemotron 3.5 Transcribe 30B-A4B", "vendor": "NVIDIA", "vendorUrl": "https://github.com/NVIDIA-NeMo/Speech", "kind": "weights", "category": "speech-to-text", "summary": "Nemotron 3.5 Transcribe is NVIDIA's open-weight speech model of 30B parameters, about 4.3B active, that transcribes and translates recordings up to an hour in 40 languages, with speaker tags and word timestamps. OpenMDW-1.1. The 4-bit build fits 32 GB.", "url": "https://www.anchorterminal.com/tools/nemotron-3-5-transcribe-30b-a4b", "markdownUrl": "https://www.anchorterminal.com/tools/nemotron-3-5-transcribe-30b-a4b.md", "slimMarkdownUrl": "https://www.anchorterminal.com/tools/nemotron-3-5-transcribe-30b-a4b.min.md", "jsonUrl": "https://www.anchorterminal.com/api/v1/tools/nemotron-3-5-transcribe-30b-a4b.json", "dns": "nemotron-3-5-transcribe-30b-a4b.grade.anchorterminal.com", "repo": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B", "license": "OpenMDW-1.1", "transports": [], "packages": [ { "registry": "huggingface", "name": "nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B" }, { "registry": "huggingface", "name": "nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B-FP8" }, { "registry": "huggingface", "name": "nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B-NVFP4" } ], "auth": "none", "authNotes": "The weights download from Hugging Face with no account, token or terms to accept. The BF16, FP8 and NVFP4 repositories are not gated (checked 2026-11-10).", "pricing": "free", "pricingNotes": "Free to download and use under OpenMDW-1.1, a permissive licence with no use policy or threshold. Running it yourself costs the hardware: 21.6 GB at 4 bits (NVFP4), 34.7 GB in FP8 or 67.5 GB in BF16. DeepInfra serves it for those without a GPU (checked 2026-11-10).", "priceSummary": "Free", "where": "local", "x402": { "level": "no", "evidence": "Weights you download and run yourself, so there is nothing to pay per call and no payment protocol (checked 2026-11-10).", "endpoints": [] }, "toolCount": null, "popularity": { "githubStars": null, "npmWeekly": null, "pypiWeekly": null, "asOf": "2026-11-10" }, "docsUrl": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B", "capabilities": [ "speech.stt", "speech.batch", "speech.languages", "speech.diarisation", "speech.translation", "model.open-weights" ], "tags": [ "open-weights", "model", "speech-to-text", "asr", "multilingual", "translation", "diarization", "timestamps", "long-form", "moe", "reasoning", "openmdw", "safetensors", "fp8", "nvfp4", "self-hosted" ], "lastRelease": "2026-10-29", "graded": true, "anchor": { "graded": true, "score": 61, "grade": "C", "agentReady": false, "rank": 0, "ranked": false, "notRankedWhy": "An open-weight model, graded by the open-weight checklist and ranked among open-weight models, not against tools", "rankOf": 1518, "categoryRank": 20, "methodology": "0.7", "run": "2026-10-01", "scores": { "interface": 85, "reach": 75, "risk": 75, "upkeep": 77 }, "pending": [ "performance", "tasks" ], "breakdown": [ { "key": "capability", "name": "Capability", "weight": 35, "effectiveWeight": 0, "pending": true, "points": 0, "reason": "Pending for this listing. Nobody independent has published results for it in at least two areas yet, so its weight is shared across the other four and the total is capped at 61 until they do or our own runs start." }, { "key": "reach", "name": "Reach", "weight": 25, "effectiveWeight": 38.46, "score": 75, "points": 28.85, "reason": "Memory 20, the smallest build is NVIDIA's NVFP4 repository, 21.56 GB, and 21.56 GB × 1.2 = 25.9 GB, within 32 GB. FP8 is 34.67 GB and BF16 67.5 GB. Runtimes 15, Released Transformers 5.22.2 has the nemotron3_5_transcribe model (5). Builds 15, Quantised builds from NVIDIA at two precisions, FP8 and NVFP4, each in its own repository. Terms 10, OpenMDW-1.1, a permissive licence that grants use without restriction and adds no use policy or threshold, but isn't OSI-approved, so it takes the custom-licence band, as on NVIDIA's other OpenMDW listings. Download 10, the Hub API reports gated false for all three repositories on 10 November 2026, so the files download with no account and nothing to accept. Hosted 5, the card gives a release on build.nvidia.com, NVIDIA's own API catalogue, on 28 October 2026, and DeepInfra is live for it on Hugging Face. 20 + 15 + 15 + 10 + 10 + 5 = 75.", "checked": "2026-11-10" }, { "key": "risk", "name": "Risk", "weight": 20, "effectiveWeight": 30.77, "score": 75, "points": 23.08, "reason": "File safety 25, safetensors only in all three repositories, no .bin, .pt or .pth files and no Python. Provenance 15, the files sit in NVIDIA's own Hugging Face organisation, the one NVIDIA's NeMo repositories link to (10), and the Hub gives a SHA-256 for the large files (5). Safety evaluation 0, no safety results. Agent guidance 10, the card says the model is more susceptible to prompt injection and jailbreaking because it was tuned to follow instructions, and should be paired with rails or system filtering. Licence risk 15, OpenMDW-1.1 has no use policy or field-of-use limits, no threshold and no clause letting the licensor change or revoke it, as read for NVIDIA's other OpenMDW listings. Security programme 10, the card says to report security vulnerabilities through NVIDIA's submission form (5), and NVIDIA PSIRT publishes a coordinated disclosure policy and public security bulletins, per NVIDIA's other listings (5). 25 + 15 + 0 + 10 + 15 + 10 = 75.", "checked": "2026-11-10" }, { "key": "interface", "name": "Agent interface", "weight": 10, "effectiveWeight": 15.38, "score": 85, "points": 13.08, "reason": "Chat template 20, chat_template.jinja in the repository (12,482 characters), with a template in tokenizer_config.json too. Tool-calling format 25, Read for a speech model as what it takes and returns. Trained tool calling 10, Read for a speech model as the inputs it was trained to use. Structured output 15, Read for a speech model as typed outputs. Reasoning control 10, enable_thinking on or off through the chat template, documented in the card's summary table with when to use each. Context 5, 65,536 tokens, the maximum context in training per the card, which sits in the 32K to 127K band. 20 + 25 + 10 + 15 + 10 + 5 = 85.", "checked": "2026-11-10" }, { "key": "upkeep", "name": "Documentation and upkeep", "weight": 10, "effectiveWeight": 15.38, "score": 77, "points": 11.85, "reason": "Model card 17, Intended use, limitations (the explainability subcard's technical limitations), the architecture, the parameters, the context and the languages are on the card. Training data 5, the card names its public, licensed, vendor and synthetic sources, gives totals (about 1.72 million hours and 5.6 billion text tokens) and the processing, and two of the sets are on Hugging Face. Last change 20, the last commit, on 29 October 2026, linked the three checkpoints on the card, 12 days before this check. Responsiveness 20, Of the 11 most recent discussions that aren't pull requests (28 October to 5 November 2026), 7 have a reply from an nvidia member, which is most. Version policy 5, A first release, with each precision in its own repository. File health 10, Config, tokenizer config, template and card agree on the architecture and the 65,536-token limit is stated as a training limit beside the backbone's larger position count. 17 + 5 + 20 + 20 + 5 + 10 = 77.", "checked": "2026-11-10" } ], "assessment": { "date": "2026-11-10", "basis": "public evidence", "confidence": "low", "notes": { "capability": "Pending for this listing. Independent results were found in 0 areas (none), and Capability needs 2. Its weight is shared across the other four categories and the total is capped at 61 until there are more, or our own runs start.", "interface": "Chat template 20, chat_template.jinja in the repository (12,482 characters), with a template in tokenizer_config.json too. Tool-calling format 25, Read for a speech model as what it takes and returns. Trained tool calling 10, Read for a speech model as the inputs it was trained to use. Structured output 15, Read for a speech model as typed outputs. Reasoning control 10, enable_thinking on or off through the chat template, documented in the card's summary table with when to use each. Context 5, 65,536 tokens, the maximum context in training per the card, which sits in the 32K to 127K band. 20 + 25 + 10 + 15 + 10 + 5 = 85.", "reach": "Memory 20, the smallest build is NVIDIA's NVFP4 repository, 21.56 GB, and 21.56 GB × 1.2 = 25.9 GB, within 32 GB. FP8 is 34.67 GB and BF16 67.5 GB. Runtimes 15, Released Transformers 5.22.2 has the nemotron3_5_transcribe model (5). Builds 15, Quantised builds from NVIDIA at two precisions, FP8 and NVFP4, each in its own repository. Terms 10, OpenMDW-1.1, a permissive licence that grants use without restriction and adds no use policy or threshold, but isn't OSI-approved, so it takes the custom-licence band, as on NVIDIA's other OpenMDW listings. Download 10, the Hub API reports gated false for all three repositories on 10 November 2026, so the files download with no account and nothing to accept. Hosted 5, the card gives a release on build.nvidia.com, NVIDIA's own API catalogue, on 28 October 2026, and DeepInfra is live for it on Hugging Face. 20 + 15 + 15 + 10 + 10 + 5 = 75.", "risk": "File safety 25, safetensors only in all three repositories, no .bin, .pt or .pth files and no Python. Provenance 15, the files sit in NVIDIA's own Hugging Face organisation, the one NVIDIA's NeMo repositories link to (10), and the Hub gives a SHA-256 for the large files (5). Safety evaluation 0, no safety results. Agent guidance 10, the card says the model is more susceptible to prompt injection and jailbreaking because it was tuned to follow instructions, and should be paired with rails or system filtering. Licence risk 15, OpenMDW-1.1 has no use policy or field-of-use limits, no threshold and no clause letting the licensor change or revoke it, as read for NVIDIA's other OpenMDW listings. Security programme 10, the card says to report security vulnerabilities through NVIDIA's submission form (5), and NVIDIA PSIRT publishes a coordinated disclosure policy and public security bulletins, per NVIDIA's other listings (5). 25 + 15 + 0 + 10 + 15 + 10 = 75.", "upkeep": "Model card 17, Intended use, limitations (the explainability subcard's technical limitations), the architecture, the parameters, the context and the languages are on the card. Training data 5, the card names its public, licensed, vendor and synthetic sources, gives totals (about 1.72 million hours and 5.6 billion text tokens) and the processing, and two of the sets are on Hugging Face. Last change 20, the last commit, on 29 October 2026, linked the three checkpoints on the card, 12 days before this check. Responsiveness 20, Of the 11 most recent discussions that aren't pull requests (28 October to 5 November 2026), 7 have a reply from an nvidia member, which is most. Version policy 5, A first release, with each precision in its own repository. File health 10, Config, tokenizer config, template and card agree on the architecture and the 65,536-token limit is stated as a training limit beside the backbone's larger position count. 17 + 5 + 20 + 20 + 5 + 10 = 77." }, "sources": [ { "what": "model card and its bias, explainability, safety and privacy subcards", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B", "seen": "2026-11-10" }, { "what": "gating, dates, parameter count, downloads, likes, file list and hashes", "url": "https://huggingface.co/api/models/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B?blobs=true", "seen": "2026-11-10" }, { "what": "the FP8 repository: size and precisions", "url": "https://huggingface.co/api/models/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B-FP8?blobs=true", "seen": "2026-11-10" }, { "what": "the NVFP4 repository: size and precisions", "url": "https://huggingface.co/api/models/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B-NVFP4?blobs=true", "seen": "2026-11-10" }, { "what": "inference providers, DeepInfra live", "url": "https://huggingface.co/api/models/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B?expand[]=inferenceProviderMapping", "seen": "2026-11-10" }, { "what": "commit history", "url": "https://huggingface.co/api/models/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B/commits/main", "seen": "2026-11-10" }, { "what": "architecture and positions", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B/raw/main/config.json", "seen": "2026-11-10" }, { "what": "chat template", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B/raw/main/chat_template.jinja", "seen": "2026-11-10" }, { "what": "discussions, replies counted by script from authors, dates and member flags", "url": "https://huggingface.co/api/models/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B/discussions", "seen": "2026-11-10" }, { "what": "the OpenMDW-1.1 licence, as read for our other NVIDIA listings", "url": "https://openmdw.ai/license/1-1/", "seen": "2026-11-10" }, { "what": "Transformers v5.22.2 nemotron3_5_transcribe", "url": "https://github.com/huggingface/transformers/tree/v5.22.2/src/transformers/models/nemotron3_5_transcribe", "seen": "2026-11-10" }, { "what": "llama.cpp b12242, the converter (conversion/nemotron.py) and the audio front end (tools/mtmd), shallow clone", "url": "https://github.com/ggml-org/llama.cpp/tree/b12242/tools/mtmd", "seen": "2026-11-10" }, { "what": "vLLM v0.33.0 registry, no nemo_speechlm", "url": "https://raw.githubusercontent.com/vllm-project/vllm/v0.33.0/vllm/model_executor/models/registry.py", "seen": "2026-11-10" }, { "what": "NVIDIA's vulnerability submission form, linked from the card", "url": "https://www.nvidia.com/en-us/support/submit-security-vulnerability/", "seen": "2026-11-10" }, { "what": "our other NVIDIA listings, for company facts", "url": "https://www.anchorterminal.com/tools/nemotron-3-5-lightning-30b-a3b", "seen": "2026-11-10" } ], "openQuestions": [ "Runtimes credits SGLang through SGLang-Omni 0.2.0, the SGLang project's released audio server. A reading that counts only SGLang itself gives 10", "llama.cpp is credited on its converter and audio front end at b12242, and NVIDIA publishes no GGUF. The card's GGML route wasn't traced to a named project", "Training data is 5 by the settled reading on cut-offs, for a card that names more sources than most. A reading by sources and sizes gives 15", "Hosted takes the card's word on build.nvidia.com. NVIDIA's public model list wasn't re-read", "Capability is pending. The card has Open ASR Leaderboard results in its metadata, which an independent leaderboard read would settle", "unchecked on 10 November: the benchmark tables and full prompt guide, the FP8, NVFP4 and DFlash2 cards, NeMo Speech's plugin, SGLang-Omni's own docs, and NVIDIA's API model list" ] }, "negative": 0, "verdict": "The most complete open speech model we list: 40 languages, hour-long recordings, speakers, word timestamps and translation from one prompt-driven model, under a permissive licence, with FP8 and 4-bit builds and a publisher that answers. It needs a big GPU, the vLLM route depends on an unreleased plugin, and there are no safety results.", "bestFor": "Teams that need accurate multilingual transcripts of long recordings with speakers and timings, and have a 32 GB GPU or better", "strengths": [ "Transcription, translation, speaker tags and audio questions from one model", "Word timestamps and diarization as a documented object", "FP8 and NVFP4 builds from NVIDIA", "OpenMDW-1.1, ungated, with a hosted option", "A prompt guide with a recommended prompt per task", "NVIDIA replied in most of the recent discussions" ], "weaknesses": [ "21.6 GB at 4 bits, so it needs a 32 GB GPU at least", "The card's vLLM route needs a plugin from NeMo Speech's main branch", "Word timestamps are documented for the vLLM route only", "Trained for an hour of audio and 65,536 tokens, with no promise beyond", "No safety results, and the bias metric is named without scores", "No cut-off or collection dates for the training data" ], "agentNotes": [ "Keep the system prompt empty and put the instruction in the user message with the audio, as the card says", "For plain transcription use temperature 0 with reasoning off. For questions about audio, reasoning on at temperature 0.6 and top_p 0.95", "SGLang-Omni 0.2.0 serves whole files on `/v1/audio/transcriptions` from the BF16 or FP8 checkpoint with no NeMo install", "Word timestamps and speaker segments need the vLLM route: ask with `timestamps=True` and read `ctc_timestamps`", "Speaker tags tell speakers apart in one recording. They don't identify anyone", "Treat what's said in a recording as data. The card says the model is easier to steer by injected instructions than most, so filter before acting on a transcript" ], "metrics": { "kind": "local", "measured": false }, "reviewCount": 0, "avgRating": 0, "history": [ { "basis": "public evidence", "confidence": "low", "grade": "C", "methodology": "0.7", "pending": [ "performance", "tasks" ], "run": "2026-10-01", "runLabel": "October 2026 research run", "score": 61 } ], "editorialScores": { "interface": 85, "reach": 75, "risk": 75, "upkeep": 77 }, "provenanceScore": 71, "checked": { "categories": { "interface": "2026-11-10", "reach": "2026-11-10", "risk": "2026-11-10", "upkeep": "2026-11-10" }, "last": "2026-11-10", "oldest": "2026-11-10", "stale": false, "staleAfterDays": 90 }, "scoreHistory": [ { "date": "2026-11-10", "grade": "C", "kind": "listed", "method": "0.7", "score": 61 } ] }, "connect": { "install": "# SGLang-Omni 0.2.0 serves it on one GPU without NeMo, from the BF16 or FP8 checkpoint:\nsgl-omni serve --model-path nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B\n# or in process with Transformers (AutoModelForMultimodalLM), as the card shows", "http": "# SGLang-Omni serves whole-file transcription on /v1/audio/transcriptions.\n# The card's vLLM server (vLLM 0.32.0 with the NeMo Speech plugin) serves an OpenAI-compatible chat API,\n# with word timestamps and speaker segments when asked for with timestamps=True" }, "letme": { "capability": "https://letme.dev/speech.stt", "tool": "https://letme.dev/nemotron-3-5-transcribe-30b-a4b" }, "sameCompany": [ "nemotron-3-5-super-vl", "nvidia-nemo-retriever", "nemo-guardrails", "nemotron-3-5-asr-streaming-0-6b", "parakeet-tdt-0-6b-v3", "parakeet-unified-en-0-6b", "nemotron-3-5-parse", "cosmos-reason2-2b", "nemotron-3-5-bolt-e4b-4gb", "nemotron-3-nano-30b-a3b", "nemotron-3-nano-4b", "nemotron-3-nano-omni-30b-a3b-reasoning", "nemotron-3-super-120b-a12b", "nemotron-3-ultra-550b-a55b", "nemotron-3-5-lightning-30b-a3b" ], "weights": { "trainedBy": "NVIDIA", "builtOn": "NVIDIA Nemotron 3.5 Lightning 30B-A3B, per the card and the metadata", "params": "30B", "active": "4.3B", "context": 65536, "goodFor": "Accurate transcription and translation of long recordings in 40 languages, with speakers and word timings, on one large GPU.", "runtimes": [ "llama.cpp", "SGLang", "Transformers" ], "builds": [ { "name": "NVIDIA-Nemotron-3.5-Transcribe-30B-A4B-NVFP4", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B-NVFP4", "who": "publisher", "precision": "nvfp4", "sizeGB": 21.56, "memoryGB": 32, "format": "safetensors", "runsOn": [ "Transformers" ], "note": "NVIDIA's 4-bit build. The card lists it for Blackwell GPUs, down to an RTX 5090" }, { "name": "NVIDIA-Nemotron-3.5-Transcribe-30B-A4B-FP8", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B-FP8", "who": "publisher", "precision": "fp8", "sizeGB": 34.67, "memoryGB": 48, "format": "safetensors", "runsOn": [ "SGLang", "Transformers" ], "note": "NVIDIA's FP8 build. SGLang-Omni serves it" }, { "name": "NVIDIA-Nemotron-3.5-Transcribe-30B-A4B (BF16)", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B", "who": "publisher", "precision": "bf16", "sizeGB": 67.5, "memoryGB": 96, "format": "safetensors", "runsOn": [ "llama.cpp", "SGLang", "Transformers" ], "note": "The reference checkpoint, for accuracy and fine-tuning. The one layout llama.cpp converts" } ], "run": "sgl-omni serve --model-path nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B-FP8", "runBuild": "NVIDIA-Nemotron-3.5-Transcribe-30B-A4B-FP8", "checked": "2026-11-10", "minMemoryGB": 32 }, "family": { "lab": "NVIDIA (Nemotron)", "family": "Nemotron 3", "url": "https://www.anchorterminal.com/open-weights/#nvidia", "others": [ { "name": "Nemotron 3 Nano 30B-A3B", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16", "params": "30B total, 3B active (31.6B per the safetensors index)", "listing": "https://www.anchorterminal.com/tools/nemotron-3-nano-30b-a3b" }, { "name": "Nemotron 3 Nano 30B-A3B Base", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-Base-BF16", "params": "30B total, 3B active" }, { "name": "Nemotron 3 Super 120B-A12B", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16", "params": "120B total, 12B active", "listing": "https://www.anchorterminal.com/tools/nemotron-3-super-120b-a12b" }, { "name": "Nemotron 3 Super 120B-A12B Base", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16", "params": "120B total, 12B active" }, { "name": "Nemotron 3 Nano 4B", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16", "params": "4B", "listing": "https://www.anchorterminal.com/tools/nemotron-3-nano-4b" }, { "name": "Nemotron 3 Nano Omni 30B-A3B Reasoning", "url": "https://huggingface.co/nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16", "params": "30B total, 3B active", "listing": "https://www.anchorterminal.com/tools/nemotron-3-nano-omni-30b-a3b-reasoning" }, { "name": "Nemotron 3 Ultra 550B-A55B", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16", "params": "550B total, 55B active", "listing": "https://www.anchorterminal.com/tools/nemotron-3-ultra-550b-a55b" }, { "name": "Nemotron 3 Ultra 550B-A55B Base", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-Base-BF16", "params": "550B total, 55B active" }, { "name": "Nemotron 3 Ultra 550B-A55B GenRM", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-GenRM", "params": "550B total, 55B active" }, { "name": "Nemotron 3.5 Lightning 30B-A3B", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16", "params": "30B total, 3B active", "listing": "https://www.anchorterminal.com/tools/nemotron-3-5-lightning-30b-a3b" }, { "name": "Nemotron 3.5 Lightning 30B-A3B Base", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-Base-BF16", "params": "30B total, 3B active" }, { "name": "Nemotron 3.5 Super VL 120B-A12B", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16", "params": "120B total, 12B active", "listing": "https://www.anchorterminal.com/tools/nemotron-3-5-super-vl" }, { "name": "Nemotron 3.5 Super VL 120B-A12B Base", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-Base-BF16", "params": "120B total, 12B active" }, { "name": "Nemotron 3.5 Bolt 74B-A9B", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF16", "params": "74B total, 9B active" }, { "name": "Nemotron 3.5 Bolt E4B 4GB", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB", "params": "8B total, 4B effective", "listing": "https://www.anchorterminal.com/tools/nemotron-3-5-bolt-e4b-4gb" }, { "name": "Nemotron Labs 3 Elastic 30B-A3B", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-3-Elastic-30B-A3B-BF16", "params": "30B total, 3B active" }, { "name": "Nemotron Labs 3 Puzzle 75B-A9B", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-3-Puzzle-75B-A9B-BF16", "params": "75B total, 9B active" }, { "name": "Nemotron 3 Labs Ultra Math SFT", "url": "https://huggingface.co/nvidia/Nemotron-3-Labs-Ultra-Math-SFT", "params": "550B total, 55B active" }, { "name": "Nemotron 3 Labs Ultra Math RL", "url": "https://huggingface.co/nvidia/Nemotron-3-Labs-Ultra-Math-RL", "params": "550B total, 55B active" }, { "name": "Nemotron Labs 3 Competitive Coding 550B-A55B", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-3-Competitive-Coding-550B-A55B-NVFP4", "params": "550B total, 55B active" }, { "name": "nemotron-3.5-asr-streaming-0.6b", "url": "https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b", "params": "0.6B (638M per the safetensors count)", "listing": "https://www.anchorterminal.com/tools/nemotron-3-5-asr-streaming-0-6b" }, { "name": "Nemotron 3.5 Parse", "url": "https://huggingface.co/nvidia/Nemotron-3.5-Parse", "params": "0.631B (630,800,496 trainable parameters per the card)", "listing": "https://www.anchorterminal.com/tools/nemotron-3-5-parse" } ], "olderCount": 47 }, "notable": [ "72,455 downloads in the last 30 days and 3,147 likes on Hugging Face on 10 November 2026, two weeks after release (https://huggingface.co/api/models/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B)", "A language model that listens: NVIDIA's Nemotron 3.5 Lightning backbone with a speech encoder and a speaker encoder in front, so the instruction in the prompt chooses verbatim or clean transcripts, translation, speaker tags or questions about the audio (https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B)", "Trained on recordings up to one hour with a context of 65,536 tokens, and the card says it isn't guaranteed beyond either (https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B)", "NVIDIA publishes FP8 and NVFP4 builds and a draft model for speculative decoding beside the BF16 weights (https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B)", "The card's vLLM route needs a plugin from the main branch of NeMo Speech. SGLang-Omni 0.2.0, released Transformers and llama.cpp run it without NeMo (https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B)", "The card says the model is more open to prompt injection because it was tuned to follow instructions, and should sit behind rails or filtering (https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B)", "An NVIDIA member replied in 7 of the 11 most recent discussions (https://huggingface.co/api/models/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B/discussions)" ], "area": "voice", "details": [ { "label": "Surface graded", "value": "The open weights at huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B and the FP8 and NVFP4 repositories beside it, read from the model card and its subcards, the files and their metadata, the config, the chat template, the commit history and the runtimes' own code at their latest releases on 10 November 2026. The card is long; the benchmark tables, the full prompt guide and the vLLM environment setup were skimmed, not read line by line. NVIDIA's company facts come from our other NVIDIA listings. Nothing was downloaded or run" }, { "label": "Model", "value": "33,697,781,029 parameters in the BF16 files. The card says about 4.3B are active, the audio encoder included. A speech-augmented language model: a Transformer speech encoder and the Nemotron-3-Diarization speaker encoder feed NVIDIA's Nemotron 3.5 Lightning 30B-A3B backbone, a hybrid Mamba-Transformer mixture of experts, with heads for word timestamps and multi-token prediction" }, { "label": "Languages", "value": "40, per the card's metadata and its language support matrix, among them English, the main European languages, Hindi, Japanese, Korean, Chinese, Arabic, Thai, Turkish and Vietnamese" }, { "label": "Files", "value": "One safetensors file in BF16 (67.5 GB), chat_template.jinja, config, processor and tokenizer files, an alignment tokenizer, the LICENSE and third-party notices. No pickle files and no Python. The FP8 repository is 34.7 GB and the NVFP4 one 21.6 GB. Hugging Face lists a SHA-256 for the large files" }, { "label": "Licence", "value": "OpenMDW-1.1, named in the metadata and on the card, with a LICENSE file in the repository. The card says the model is ready for commercial use. Third-party notices cover components from Qwen3-ASR and vLLM" }, { "label": "Runtimes", "value": "The card documents vLLM 0.32.0 with a plugin from NeMo Speech's main branch, SGLang-Omni 0.2.0 on SGLang 0.5.22 (whole-file transcription, no NeMo), Transformers, and GGML on some GPUs. Released Transformers 5.22.2 has the model, and llama.cpp b12242 has its converter and audio front end, for the BF16 layout only. Released vLLM 0.33.0 has no code for it without the plugin" }, { "label": "Input and output", "value": "Audio (anything that decodes to 16 kHz mono PCM) with a text instruction and optional context such as vocabulary hints. Text out: transcripts, translations, speaker and event tags, language labels, answers, or PII annotations, depending on the prompt. With timestamp capture on in vLLM the response adds a ctc_timestamps object with each word's start and end, speaker tags and diarization segments" }, { "label": "Hardware", "value": "The card's quick start serves the BF16 checkpoint on one H100 80 GB. Its hardware table lists NVFP4 on an RTX 5090 and all three precisions on Blackwell servers, DGX Spark and Jetson Thor" }, { "label": "Training data", "value": "About 1.72 million hours of unique indexed audio and 5.6 billion unique text tokens, per the card, from named public sets (LibriSpeech, VoxPopuli, Common Voice 26, YODAS2, Granary v2 and others), licensed corpora such as Fisher and Switchboard, vendor data and NVIDIA's synthetic sets. Some sources are web-derived. No cut-off date" }, { "label": "Reasoning", "value": "enable_thinking in the chat template, on or off. The card recommends reasoning off at temperature 0 for transcription, translation and language identification, and on at temperature 0.6 and top_p 0.95 for questions about audio" } ], "provenance": { "legalEntity": "NVIDIA Corporation", "domain": "nvidia.com", "domainRegistered": "1993-04-20", "endpointOnVendorDomain": null, "terms": "https://openmdw.ai/license/1-1/", "privacy": "", "statusPage": "", "changelog": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B/commits/main", "securityTxt": "none", "checked": "2026-11-10", "notes": [ "The model sits in NVIDIA's Hugging Face organisation and the card's governing terms are OpenMDW-1.1. The domain registration date is carried from the RDAP check made for nemotron-3-nano-30b-a3b on 21 October 2026.", "Nothing is hosted under this listing, so the licence stands in for terms of service and no privacy policy is scored.", "The changelog is the commit history, 13 commits from 27 to 29 October 2026.", "security.txt is carried from our other NVIDIA listings and wasn't re-read." ], "score": 71, "checks": [ { "check": "Legal entity named", "value": "NVIDIA Corporation", "points": 20, "max": 20, "state": "ok" }, { "check": "Domain age", "value": "nvidia.com, registered 1993-04-20 (33 years)", "points": 15, "max": 15, "state": "ok" }, { "check": "Endpoint on the vendor's domain", "value": "no hosted endpoint", "points": 0, "max": 0, "state": "na" }, { "check": "Terms of service", "value": "published, not read by us yet", "points": 4, "max": 10, "state": "part" }, { "check": "Privacy policy", "value": "nothing hosted, not scored", "points": 0, "max": 0, "state": "na" }, { "check": "Status page", "value": "not found", "points": 0, "max": 10, "state": "no" }, { "check": "Changelog", "value": "published", "points": 10, "max": 10, "state": "ok" }, { "check": "security.txt", "value": "not found", "points": 0, "max": 10, "state": "no" }, { "check": "People named", "value": "4 named with their role, 1 with a public profile", "points": 10, "max": 10, "state": "ok" }, { "check": "Support route", "value": "help pages, a community", "points": 5, "max": 5, "state": "ok" } ] }, "humans": { "slug": "nemotron-3-5-transcribe-30b-a4b", "checked": "2026-11-10", "company": "NVIDIA", "people": [ { "name": "Jensen Huang", "role": "Founder, president and CEO, NVIDIA", "founder": true, "photo": "/assets/img/people/nvidia-jensen-huang.webp", "photoSource": "https://nvidianews.nvidia.com/bios", "bio": "Jensen Huang founded NVIDIA in 1993 and has been its president, chief executive and a board member since its start. Before NVIDIA he worked at LSI Logic and Advanced Micro Devices. He holds electrical engineering degrees from Oregon State University and Stanford University and is a member of the National Academy of Engineering.", "bioSources": [ "https://nvidianews.nvidia.com/bios/jensen-huang" ], "page": "/people/jensen-huang/" }, { "name": "Chris Malachowsky", "role": "Co-founder and NVIDIA Fellow, NVIDIA", "founder": true, "photo": "/assets/img/people/nvidia-chris-malachowsky.webp", "photoSource": "https://nvidianews.nvidia.com/bios", "bio": "Chris Malachowsky founded NVIDIA in 1993 with Jensen Huang and Curtis Priem and is an NVIDIA Fellow on its executive staff. He has led its IT, operations and product engineering and, most recently, its research organisation. He has authored close to 40 patents and was inducted into the Florida Inventors Hall of Fame in 2019.", "bioSources": [ "https://nvidianews.nvidia.com/bios/chris-a-malachowsky", "https://en.wikipedia.org/wiki/Curtis_Priem" ], "page": "/people/chris-malachowsky/" }, { "name": "Curtis Priem", "role": "Co-founder, NVIDIA; left in 2003", "founder": true, "bio": "Curtis Priem co-founded NVIDIA in 1993 and was its chief technical officer until 2003. He earlier designed the IBM Professional Graphics Adapter at Vermont Microsystems and developed the GX graphics chip as a senior staff engineer at Sun Microsystems.", "bioSources": [ "https://en.wikipedia.org/wiki/Curtis_Priem", "https://en.wikipedia.org/wiki/Nvidia" ], "page": "/people/curtis-priem/" }, { "name": "Nithin Rao Koluguri", "role": "Senior Research Scientist; Parakeet-TDT-0.6B-v3 paper co-author", "linkedin": "https://www.linkedin.com/in/nithinraok", "page": "/people/nithin-rao-koluguri/" } ], "peopleNote": "NVIDIA's executive bios page lists Jensen Huang (Founder \u0026 CEO) and Chris A. Malachowsky (Founder \u0026 NVIDIA Fellow) as its founders. Curtis Priem, its third co-founder, was chief technical officer until he left in 2003, per Wikipedia. Product people on single listings (Nemotron, NeMo Guardrails, Parakeet) are kept as they are.", "companySocial": { "linkedin": "https://www.linkedin.com/company/nvidia", "x": "https://x.com/nvidia", "github": "https://github.com/NVIDIA" }, "support": { "urls": [ "https://github.com/NVIDIA-NeMo/Speech/issues", "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B/discussions" ], "community": [ { "label": "NeMo Speech discussions", "url": "https://github.com/NVIDIA-NeMo/Speech/discussions", "kind": "github" }, { "label": "Model community", "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B/discussions", "kind": "link" } ] }, "sources": [ "https://arxiv.org/abs/2509.14128", "https://www.linkedin.com/in/nithinraok/details/experience/", "https://www.linkedin.com/in/nithinraok", "https://developer.nvidia.com/nemotron", "https://github.com/NVIDIA-NeMo/Speech", "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B", "https://nvidianews.nvidia.com/bios", "https://nvidianews.nvidia.com/bios/jensen-huang", "https://nvidianews.nvidia.com/bios/chris-a-malachowsky", "https://en.wikipedia.org/wiki/Nvidia", "https://en.wikipedia.org/wiki/Curtis_Priem" ], "ownership": { "status": "public", "exchange": "Nasdaq", "ticker": "NVDA", "checked": "2026-10-28", "sources": [ "https://nvidianews.nvidia.com/news/nvidia-and-oracle-expand-nvidia-vera-rubin-to-multi-gigawatt-scale-and-bring-nvidia-vera-cpu-to-oracle-cloud-infrastructure", "https://nvidianews.nvidia.com/bios" ] } }, "pageJsonUrl": "https://www.anchorterminal.com/tools/nemotron-3-5-transcribe-30b-a4b.json" } }