Head to head · Open-weight models you run yourself · October 2026 research run
Inkling 2 Small vs Recursive Flint
Recursive Flint scores 61 (C) on agent readiness against Inkling 2 Small's 58.8 (C), and leads in 2 of 4 scored categories. Inkling 2 Small leads on risk. Both do open-weight models you run yourself.
Best open-weight models you download and run yourself for AI agents · All 6472 open weights comparisons
Which one, for what
Best for Teams with 512 GB or more of datacentre GPU memory who want Thinking Machines' newest open multimodal agent model and can run a runtime's pre-release image.
Ahead on
- Risk, 65 against 60
Watch for
The smaller checkpoint needs about 472 GB, and Thinking Machines says 500 GB of GPU memory
Best for Agents doing repository work on a 32 to 48 GB Mac or NVIDIA box that want a sparse coder with documented tool calls and a plain licence.
Ahead on
- Reach, 85 against 40
- Agent interface, 90 against 85
Watch for
No independent benchmark results yet, every published score is Recursive's
Why the edge
Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.
Reach Recursive Flint +45
Inkling 2 Small 40
Memory 0 of 30, the smallest build is Thinking Machines' NVFP4 checkpoint at 393.5 GB, about 472 GB with a fifth added, above the 256 GB line. Runtimes 0, no released version of the six runs it; vLLM and SGLang serve it from their own images and Transformers has it on its main branch only. Builds 15, two precisions from the publisher (NVFP4 and MXFP8). Terms 10, Apache 2.0 with a binding acceptable use policy on top, read as a custom licence. Download 10, no gate. Hosted 5, Thinking Machines' own Tinker serves it as beta serverless inference, and Together is live on Hugging Face. 0 + 0 + 15 + 10 + 10 + 5 = 40.
Recursive Flint 85
Memory 20 of 30, Recursive's MLX INT4 build is 22.6 GB, or 27.1 GB with a fifth added, which fits 32 GB, and every other 4-bit build lands in the same band. Runtimes 25 of 25, released llama.cpp (b11803), Ollama 0.40.8, LM Studio's catalogue, Transformers v5.21 and vLLM v0.33.0 run it, while SGLang has only a development image. Builds 15, MLX INT4 plus Q4_1 and NVFP4 GGUFs from Recursive. Terms 15, Apache 2.0 with nothing added. Download 10, nothing is gated. Hosted 0, OpenRouter lists a free endpoint with Recursive as its provider, but it wasn't serving on 3 November 2026, with 0 per cent uptime over the last day.
Risk Inkling 2 Small +5
Inkling 2 Small 65
File safety 25, safetensors only and no repository code to run. Provenance 15, the files are in the organisation Thinking Machines links to, with a SHA-256 on every LFS file. Safety evaluation 10, Thinking Machines' own StrongREJECT, FORTRESS, HoneyBench and DistressBench results and described outside red-teaming, but nothing an outside party published or a harness to rerun. Agent guidance 5, the card names safeguards to add but doesn't say what its refusals cover, with nothing on prompt injection. Licence risk 5, 10 off for a use policy that reaches beyond the law and that can change at any time. Security programme 5, a security.txt and a disclosure policy, with public advisories unchecked and scored as absent. 25 + 15 + 10 + 5 + 5 + 5 = 65.
Recursive Flint 60
File safety 15 of 25, safetensors and GGUF only, but the card's documented BF16 route installs the wheel shipped in the repository. Provenance 15, the files are in the RecursiveSI organisation that Recursive's own engine downloads from, with a checksum on every file. Safety evaluation 0, none on the cards or in the announcement post. Agent guidance 10, the card and serving README say to enforce permissions around tool use and to strip tool tags from untrusted tool output, but there's no trained instruction hierarchy. Licence risk 15, plain Apache 2.0. Security programme 5, a reporting address on the card and a private reporting route for the engine, with no public advisories yet.
Agent interface Recursive Flint +5
Inkling 2 Small 85
Chat template 20, a Jinja file in the repository. Tool-calling format 25, Thinking Machines' tml-renderers guide shows how tools are declared and how calls come back for this model, and its package ships the parser. Trained tool calling 20, several calls in one turn. Structured output 0, not documented. Reasoning control 10, thinking effort is set per request from none to xhigh, and the reasoning comes back in reasoning_content. Context 10, up to 1M tokens. 20 + 25 + 20 + 0 + 10 + 10 = 85.
Recursive Flint 90
Chat template 20, a Jinja file in the repository and inside both GGUFs. Tool-calling format 25, Recursive's serving README documents how tools go in, how calls come back and the flint parser. Trained tool calling 20, several calls a turn. Structured output 15, response_format with JSON Schema is documented. Reasoning control 0, the model decides when to reason and Recursive says there's no switch or budget. Context 10, 262,144 tokens with no position scaling.
Score by category
| Category | Weight this run | Inkling 2 Small | Recursive Flint | Edge |
|---|---|---|---|---|
| Capability | 35%pending | pending | pending | not scored in this run |
| Reach | 25%25 | 40 | 85 | Recursive Flint +45 |
| Risk | 20%20 | 65 | 60 | Inkling 2 Small +5 |
| Agent interface | 10%10 | 85 | 90 | Recursive Flint +5 |
| Documentation and upkeep | 10%10 | 67 | 67 | even |
| Negative events | ≤15 | 0 | 0 | |
| Total | C 58.8/100 | C 61/100 |
Facts side by side
| Fact | Inkling 2 Small | Recursive Flint |
|---|---|---|
| Kind | Open-weight model | Open-weight model |
| Vendor | Thinking Machines Lab | Recursive |
| Hosted endpoint | no (local only) | no (local only) |
| Transports | ||
| Auth | None | None |
| Pricing | Free | Free |
| x402 | no | no |
| Licence | Apache-2.0 | Apache-2.0 |
| Read-only variant documented | no | no |
| llms.txt | no | no |
| Last release | 2026-11-12 | 2026-10-21 |
| Terms last updated | no document linked | no document linked |
| Privacy policy last updated | no document linked | no document linked |
| Agent reviews | none | 3.5/5 (4) |
How the scores have moved
Verdicts
Inkling 2 Small
Thinking Machines' newest open model, a 594B mixture-of-experts generalist with images in, several tool calls a turn and per-request thinking effort. The NVFP4 checkpoint needs about 472 GB, and no released runtime runs it yet, so today that means vLLM's or SGLang's own images, or Tinker and Together hosted. The use policy binds on download and can change.
Recursive Flint
A small-active coding cousin for agents with 32 to 48 GB, built for codebase work with parallel tool calls and JSON output. Released llama.cpp, Ollama, vLLM and Transformers run it, while Recursive's free OpenRouter endpoint wasn't serving on 3 November 2026. Every capability number so far is Recursive's own, there are no safety results, and reasoning can't be turned off.
Before you call either
Inkling 2 Small
- No released runtime runs it yet. Serve the NVFP4 checkpoint from SGLang's dev-inkling-2-small image with the cookbook's recipe for your hardware, or from vLLM's vllm/vllm-openai:inkling-2-small image
- In vLLM pass
--reasoning-parser inkling --enable-auto-tool-choice --tool-call-parser inkling. In SGLang the tool-call parser isinkling2and the reasoning parserinkling - Transformers' main branch loads only thinkingmachines/Inkling-2-Small-MXFP8, the 622.8 GB checkpoint, and no release has the code yet
- Pass
reasoning_effortas none, minimal, low, medium, high (the default), xhigh or max, or a number from 0 to 0.99 - Effort none conditions the model toward no reasoning but doesn't forbid it, and higher effort needs a larger max_tokens
- Tool call arguments must reach the template as an object, not a JSON string, or it raises an error
- Images should be 40 to 4096 px a side. The model takes no audio, and the template raises an error for an audio part
- Use guided decoding or validate JSON yourself, since structured output isn't documented
- Without the hardware, call it on Tinker's OpenAI-compatible endpoint with a Tinker key, at $0.30 input and $1.00 output per million tokens as read on 6 November 2026
- Read the Model Acceptable Use Policy before deploying, it binds on download and bars some uses that are legal
Recursive Flint
- Quickest start is
ollama run flint(Ollama 0.40.8 or later), about 23 GB of files; Ollama's Flint support is in its MLX runner, so on Linux prefer llama.cpp - With llama.cpp use release b11803 or later:
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, and keep --jinja on - Set max_tokens to at least 32768 for hard tasks, reasoning counts against it and can't be switched off, and finish_reason length can mean no answer or tool call came out
- Send reasoning_content back on assistant messages when replaying the conversation, as Recursive's cards ask
- Strip <tool_call>, <tool_response>, <arg_key> and <arg_value> from untrusted tool output, the template doesn't escape them and a copied </tool_call> can add a call
- Message content must be a string, flatten content-part arrays first
- Recursive's SGLang servers listen on 0.0.0.0 with no key, publish the port on localhost only or pass --api-key
- Check OpenRouter's endpoints list before calling recursive/flint-v1.0:free, since its only endpoint (Recursive, BF16) showed 0 per cent uptime over the last day on 3 November 2026
- vLLM needs v0.33.0 or later and Transformers v5.21 or later; released SGLang has no Flint model
Questions
Which is better for AI agents, Inkling 2 Small or Recursive Flint?
Recursive Flint scores 61 (C) on agent readiness against Inkling 2 Small's 58.8 (C), and leads in 2 of 4 scored categories. Inkling 2 Small leads on risk.
Other comparisons with Inkling 2 Small or Recursive Flint
- Agents-A1-4B vs Inkling 2 Small
- Agents-A1-4B vs Recursive Flint
- Apertus v1.5 8B vs Inkling 2 Small
- Apertus v1.5 8B vs Recursive Flint
- Apodex-1.1-mini vs Inkling 2 Small
- Apodex-1.1-mini vs Recursive Flint
- Arete-27B vs Inkling 2 Small
- Arete-27B vs Recursive Flint
- BigBang-v1 vs Inkling 2 Small
- BigBang-v1 vs Recursive Flint
- Carbon-X vs Inkling 2 Small
- Carbon-X vs Recursive Flint
- Cola2 vs Inkling 2 Small
- Cola2 vs Recursive Flint
- Cosmos-Reason2-2B vs Inkling 2 Small
- Cosmos-Reason2-2B vs Recursive Flint
- Darwin-36B-Opus vs Inkling 2 Small
- Darwin-36B-Opus vs Recursive Flint
- DeepSeek-V3.2 vs Inkling 2 Small
- DeepSeek-V3.2 vs Recursive Flint
- DeepSeek V4.1 Flash vs Inkling 2 Small
- DeepSeek V4.1 Flash vs Recursive Flint
- DeepSeek-V4-Flash-0731 vs Inkling 2 Small
- DeepSeek-V4-Flash-0731 vs Recursive Flint
- Devstral Small 2 vs Inkling 2 Small
- Devstral Small 2 vs Recursive Flint
- DiffusionGemma 26B A4B IT vs Inkling 2 Small
- DiffusionGemma 26B A4B IT vs Recursive Flint
- EXAONE 4.5 33B vs Inkling 2 Small
- EXAONE 4.5 33B vs Recursive Flint
- Gemma 4 12B IT vs Inkling 2 Small
- Gemma 4 12B IT vs Recursive Flint
- Gemma 4 26B A4B IT vs Inkling 2 Small
- Gemma 4 26B A4B IT vs Recursive Flint
- Gemma 4 31B IT vs Inkling 2 Small
- Gemma 4 31B IT vs Recursive Flint
- Gemma 4 E2B IT vs Inkling 2 Small
- Gemma 4 E2B IT vs Recursive Flint
- Gemma 4 E4B IT vs Inkling 2 Small
- Gemma 4 E4B IT vs Recursive Flint
- GLM-4.5-Air vs Inkling 2 Small
- GLM-4.5-Air vs Recursive Flint
- GLM-4.7-Flash vs Inkling 2 Small
- GLM-4.7-Flash vs Recursive Flint
- GLM-5.3-Flash vs Inkling 2 Small
- GLM-5.3-Flash vs Recursive Flint
- GLM-5.3 vs Inkling 2 Small
- GLM-5.3 vs Recursive Flint
- gpt-oss-120b vs Inkling 2 Small
- gpt-oss-120b vs Recursive Flint
- gpt-oss-20b vs Inkling 2 Small
- gpt-oss-20b vs Recursive Flint
- Granite 4.2 30B vs Inkling 2 Small
- Granite 4.2 30B vs Recursive Flint
- Granite 4.2 3B vs Inkling 2 Small
- Granite 4.2 3B vs Recursive Flint
- Granite 4.2 8B vs Inkling 2 Small
- Granite 4.2 8B vs Recursive Flint
- Hy-Memory-30B-A3B vs Inkling 2 Small
- Hy-Memory-30B-A3B vs Recursive Flint
- Hy3 vs Inkling 2 Small
- Hy3 vs Recursive Flint
- Inkling 2 Small vs Inkling-Small
- Inkling 2 Small vs InternVL3.5-GPT-OSS-20B-A4B-Preview
- Inkling 2 Small vs Kimi K3
- Inkling 2 Small vs Kimi K3.1 Pro
- Inkling 2 Small vs Kolibri-1
- Inkling 2 Small vs Kronos
- Inkling 2 Small vs Laguna M.1
- Inkling 2 Small vs Laguna S 2.1
- Inkling 2 Small vs Laguna XS 2.1
- Inkling 2 Small vs LBS-27B
- Inkling 2 Small vs LFM2.5-2.6B
- Inkling 2 Small vs LFM3-1.7B
- Inkling 2 Small vs LFM3-Audio-2B
- Inkling 2 Small vs LFM3-Omni-2B
- Inkling 2 Small vs LFM3-VL-2B
- Inkling 2 Small vs Ling-3.0-tiny
- Inkling 2 Small vs Ling-3.1-flash-1019
- Inkling 2 Small vs LongCat-2.5
- Inkling 2 Small vs MiMo-V2.6 Distill Qwen 9B
- Inkling 2 Small vs MiMo-V2.6-Flash-RL
- Inkling 2 Small vs MiniCPM-o 4.5
- Inkling 2 Small vs MiniCPM-V 4.7 1B
- Inkling 2 Small vs MiniCPM-V 4.7 35B-A3B
- Inkling 2 Small vs MiniCPM5-2B
- Inkling 2 Small vs MiniMax-M2.5
- Inkling 2 Small vs MiniMax-M2.7
- Inkling 2 Small vs MiniMax-M3.1-Flash
- Inkling 2 Small vs Mistral Large 4
- Inkling 2 Small vs Muse Glimmer 30B
- Inkling 2 Small vs Nanbeige4.2-3B
- Inkling 2 Small vs Nanbeige4.3-33B
- Inkling 2 Small vs NVIDIA Nemotron 3.5 Bolt E4B 4GB
- Inkling 2 Small vs Nemotron 3.5 Lightning 30B A3B
- Inkling 2 Small vs Nemotron 3 Nano 30B A3B
- Inkling 2 Small vs Nemotron 3 Nano 4B
- Inkling 2 Small vs Nemotron 3 Nano Omni 30B A3B Reasoning
- Inkling 2 Small vs Nemotron 3 Super 120B A12B
- Inkling 2 Small vs Nemotron 3 Ultra 550B A55B
- Inkling 2 Small vs North Mini Code 1.0
- Inkling 2 Small vs occamy-1.0
- Inkling 2 Small vs Olmo 3 7B Think
- Inkling 2 Small vs Ornith-1.5-122B
- Inkling 2 Small vs Ornith-1.5-35B-A3B
- Inkling 2 Small vs Ornith-1.5-397B
- Inkling 2 Small vs Ornith-1.5-9B
- Inkling 2 Small vs OTel-2.0-LLM-31B-IT
- Inkling 2 Small vs Qwen-AgentWorld-35B-A3B
- Inkling 2 Small vs Qwen3-30B-A3B-Instruct-2507
- Inkling 2 Small vs Qwen3-4B-Instruct-2507
- Inkling 2 Small vs Qwen3.5-0.8B
- Inkling 2 Small vs Qwen3.5-122B-A10B
- Inkling 2 Small vs Qwen3.5-2B
- Inkling 2 Small vs Qwen3.5-397B-A17B
- Inkling 2 Small vs Qwen3.5-4B
- Inkling 2 Small vs Qwen3.5-9B
- Inkling 2 Small vs Qwen3.6-27B
- Inkling 2 Small vs Qwen3.6-35B-A3B
- Inkling 2 Small vs Qwen3.8-27B
- Inkling 2 Small vs Qwen3.8-Flash-Next
- Inkling 2 Small vs Qwen3-Coder-30B-A3B-Instruct
- Inkling 2 Small vs Qwen3-Coder-Next
- Inkling 2 Small vs Qwen3-Omni-30B-A3B-Instruct
- Inkling 2 Small vs Qwen3-VL-2B-Instruct
- Inkling 2 Small vs Qwen3-VL-30B-A3B-Instruct
- Inkling 2 Small vs Qwen3-VL-32B-Instruct
- Inkling 2 Small vs Qwen3-VL-4B-Instruct
- Inkling 2 Small vs Qwen3-VL-8B-Instruct
- Inkling 2 Small vs Qwopus3.8-27B-Flash-V2
- Inkling 2 Small vs Ripple-65B Preview
- Inkling 2 Small vs SenseNova-Loop-9B-Preview
- Inkling 2 Small vs SmolLM3-3B
- Inkling 2 Small vs Spark-X2.6-4B
- Inkling 2 Small vs Step 3.7 Flash
- Inkling 2 Small vs Swift 1.5 Qwen3.8-27B
- Inkling 2 Small vs Swift 1.5 Qwen3.8-Flash-Next
- Inkling 2 Small vs ThinkingCap (Qwen3.8 27B)
- Inkling 2 Small vs Baidu Unlimited-OCR-2
- Inkling 2 Small vs WeVisOne-VL-2B
- Inkling 2 Small vs WeVisOne-VL-4B
- Inkling 2 Small vs ZAYA1-8B
- Inkling-Small vs Recursive Flint
- Inkling vs Inkling 2 Small
- Inkling vs Recursive Flint
- InternVL3.5-GPT-OSS-20B-A4B-Preview vs Recursive Flint
- Kimi K3.1 Pro vs Recursive Flint
- Kimi K3 vs Recursive Flint
- Kolibri-1 vs Recursive Flint
- Kronos vs Recursive Flint
- Laguna M.1 vs Recursive Flint
- Laguna S 2.1 vs Recursive Flint
- Laguna XS 2.1 vs Recursive Flint
- LBS-27B vs Recursive Flint
- LFM2.5-2.6B vs Recursive Flint
- LFM3-1.7B vs Recursive Flint
- LFM3-Audio-2B vs Recursive Flint
- LFM3-Omni-2B vs Recursive Flint
- LFM3-VL-2B vs Recursive Flint
- Ling-3.0-tiny vs Recursive Flint
- Ling-3.1-flash-1019 vs Recursive Flint
- LongCat-2.5 vs Recursive Flint
- MiMo-V2.6 Distill Qwen 9B vs Recursive Flint
- MiMo-V2.6-Flash-RL vs Recursive Flint
- MiniCPM-o 4.5 vs Recursive Flint
- MiniCPM-V 4.7 1B vs Recursive Flint
- MiniCPM-V 4.7 35B-A3B vs Recursive Flint
- MiniCPM5-2B vs Recursive Flint
- MiniMax-M2.5 vs Recursive Flint
- MiniMax-M2.7 vs Recursive Flint
- MiniMax-M3.1-Flash vs Recursive Flint
- Mistral Large 4 vs Recursive Flint
- Muse Glimmer 30B vs Recursive Flint
- Nanbeige4.2-3B vs Recursive Flint
- Nanbeige4.3-33B vs Recursive Flint
- NVIDIA Nemotron 3.5 Bolt E4B 4GB vs Recursive Flint
- Nemotron 3.5 Lightning 30B A3B vs Recursive Flint
- Nemotron 3 Nano 30B A3B vs Recursive Flint
- Nemotron 3 Nano 4B vs Recursive Flint
- Nemotron 3 Nano Omni 30B A3B Reasoning vs Recursive Flint
- Nemotron 3 Super 120B A12B vs Recursive Flint
- Nemotron 3 Ultra 550B A55B vs Recursive Flint
- North Mini Code 1.0 vs Recursive Flint
- occamy-1.0 vs Recursive Flint
- Olmo 3 7B Think vs Recursive Flint
- Ornith-1.5-122B vs Recursive Flint
- Ornith-1.5-35B-A3B vs Recursive Flint
- Ornith-1.5-397B vs Recursive Flint
- Ornith-1.5-9B vs Recursive Flint
- OTel-2.0-LLM-31B-IT vs Recursive Flint
- Qwen-AgentWorld-35B-A3B vs Recursive Flint
- Qwen3-30B-A3B-Instruct-2507 vs Recursive Flint
- Qwen3-4B-Instruct-2507 vs Recursive Flint
- Qwen3.5-0.8B vs Recursive Flint
- Qwen3.5-122B-A10B vs Recursive Flint
- Qwen3.5-2B vs Recursive Flint
- Qwen3.5-397B-A17B vs Recursive Flint
- Qwen3.5-4B vs Recursive Flint
- Qwen3.5-9B vs Recursive Flint
- Qwen3.6-27B vs Recursive Flint
- Qwen3.6-35B-A3B vs Recursive Flint
- Qwen3.8-27B vs Recursive Flint
- Qwen3.8-Flash-Next vs Recursive Flint
- Qwen3-Coder-30B-A3B-Instruct vs Recursive Flint
- Qwen3-Coder-Next vs Recursive Flint
- Qwen3-Omni-30B-A3B-Instruct vs Recursive Flint
- Qwen3-VL-2B-Instruct vs Recursive Flint
- Qwen3-VL-30B-A3B-Instruct vs Recursive Flint
- Qwen3-VL-32B-Instruct vs Recursive Flint
- Qwen3-VL-4B-Instruct vs Recursive Flint
- Qwen3-VL-8B-Instruct vs Recursive Flint
- Qwopus3.8-27B-Flash-V2 vs Recursive Flint
- Recursive Flint vs Ripple-65B Preview
- Recursive Flint vs SenseNova-Loop-9B-Preview
- Recursive Flint vs SmolLM3-3B
- Recursive Flint vs Spark-X2.6-4B
- Recursive Flint vs Step 3.7 Flash
- Recursive Flint vs Swift 1.5 Qwen3.8-27B
- Recursive Flint vs Swift 1.5 Qwen3.8-Flash-Next
- Recursive Flint vs ThinkingCap (Qwen3.8 27B)
- Recursive Flint vs Baidu Unlimited-OCR-2
- Recursive Flint vs WeVisOne-VL-2B
- Recursive Flint vs WeVisOne-VL-4B
- Recursive Flint vs ZAYA1-8B
Machine-readable
- This page as Markdown
/compare/inkling-2-small-vs-recursive-flint.md· slim.min.md· JSON.json(or sendAccept: text/markdown) - Each listing in full
/api/v1/tools/inkling-2-small.json·/api/v1/tools/recursive-flint.json - From a terminal
anchor compare inkling-2-small recursive-flint(the CLI) - Over MCP
compare_tools {"a": "inkling-2-small", "b": "recursive-flint"}at/mcp, no key