# MiniMax-M2.7 (slim)

> MiniMax-M2.7 is MiniMax's open-weight mixture-of-experts model for coding, agent harnesses and office work, about 229B parameters in FP8 with thinking always on, published on Hugging Face on 9 April 2026 under a non-commercial licence.

- Full: https://www.anchorterminal.com/tools/minimax-m2-7.md (~11,700 tokens) · this version ~2,580 tokens · JSON https://www.anchorterminal.com/tools/minimax-m2-7.json · canonical https://www.anchorterminal.com/tools/minimax-m2-7
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-10-21 · pricing https://www.anchorterminal.com/tools/minimax-m2-7#pricing · reviews https://www.anchorterminal.com/tools/minimax-m2-7#reviews
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

**C · 56.1/100 · rank #76 of 100 open-weight models · #61 in Open-weight models · not agent-ready · confidence high**


Assessment: A 230B mixture-of-experts model for coding and agent harnesses, with parallel tool calls, about 200K of context and parsers in vLLM and SGLang. It needs a 4 GPU server for FP8 or about 166 GB for a 4-bit library build, and it always thinks. The main risk is the licence, switched to non-commercial two days after release.

## Facts

- Kind: Open-weight model · vendor: MiniMax · category: Open-weight models · legal entity: MiniMax (named in the licence's copyright line; MiniMax's platform terms name Nanonoble Pte. Ltd., Singapore) · provenance 55/100
- Packages: huggingface `MiniMaxAI/MiniMax-M2.7`
- Auth: None · pricing: Free · x402: no · licence: MiniMax M2.7 non-commercial licence (MIT-style terms, commercial use needs MiniMax's written authorisation)
- Probe metrics: not measured yet (probes haven't run)
- Surface graded: The open weights at huggingface.co/MiniMaxAI/MiniMax-M2.7, read from the model card, the files and their metadata, the commit history, the chat template, the deployment and tool-calling guides, the licence as it is today and at each earlier revision, MiniMax's GitHub repository MiniMax-AI/MiniMax-M2.7 (raw files only) and the runtimes' own code and catalogues on 21 October 2026. Nothing was downloaded or run. MiniMax's hosted API is its own listing
- Model: A sparse mixture of experts, MiniMaxM2ForCausalLM, with 62 layers, 256 experts of which 8 are routed per token, and 3 multi-token prediction modules. 228,689,764,864 parameters in the files per Hugging Face's metadata. The card states neither the parameter count nor the architecture; LM Studio's catalogue says 230B total and 10B active. Text in, text out
- Files: 125 safetensors shards in F8_E4M3 with BF16 and F32 for the rest, 230.2 GB in all. No pickle files. Hugging Face lists a SHA-256 for every shard. The repository also holds Python model code (modeling_minimax_m2.py, configuration_minimax_m2.py) named in config.json's auto_map, and the repository is tagged custom_code
- Licence: Today a non-commercial licence on MIT-style terms. Commercial use (paid products, commercial APIs, serving fine-tuned derivatives commercially) needs MiniMax's prior written authorisation, and commercial users must show "Built with MiniMax M2.7". Personal use including self-hosted agent and app development, and non-profit or academic research, are free. A Prohibited Uses appendix bans military use among others. At release on 9 April 2026 the repository carried the MiniMax Model License instead (royalty-free, commercial use allowed, Singapore law); it was replaced on 11 April and amended on 14 April without a dated note. The card's licence link to GitHub answers with the same text
- Runtimes: MiniMax's guides for vLLM, SGLang (0.5.4.post1 or later) and Transformers 4.57.1, plus NVIDIA NIM and ModelScope. Model code in released vLLM (v0.32.0, with minimax_m2 tool and reasoning parsers), SGLang (v0.5.22), llama.cpp (minimax-m2 at b11768), Transformers 5.20.0 (minimax_m2) and mlx-lm. LM Studio's catalogue has a GGUF entry. Ollama's entry is cloud only
- Context: 204,800 tokens in config.json since a change on 20 April 2026 (196,608 at release), with no rope scaling. MiniMax's vLLM and SGLang guides still say the most per sequence is 196K. The card itself doesn't state it
- Tools and reasoning: Tools are declared as JSON Schema in `<tools>` and calls come back as `<invoke name=...>` with `<parameter name=...>` inside `<minimax:tool_call>`, several per block, documented in MiniMax's tool-calling guide with a parsing function. Parsers are minimax_m2 in vLLM and minimax-m2 in SGLang. The template opens every reply with `<think>` and has no switch to turn thinking off; earlier reasoning is kept after the last user turn. Recommended sampling `temperature=1.0`, `top_p=0.95`, `top_k=40`
- Hardware: MiniMax's guides state 220 GB for the weights and 240 GB per 1M tokens of KV cache, GPUs of compute capability 7.0 or higher, and give 4 × 96 GB GPUs (400K tokens of cache in all) and 8 × 144 GB GPUs (3M) as configurations. The smallest 4-bit library build, LM Studio's Q4_K_M GGUF, is 138.3 GB. Nothing fits in 128 GB
- Training data: Not described. The card and MiniMax's blog describe a self-evolution process in which an internal version of the model ran its own reinforcement-learning experiments and scaffold optimisation, with no sources, sizes or knowledge cut-off
- Status: MiniMax-M3 (2 June 2026) is newer, in its own repository. MiniMax's API closed M2.7 to new users on 19 October 2026 and retires it on 18 November 2026; its model list already marks M2.7 as retired. The weights stay up

- Scores: Capability 71, Reach 60, Risk 30, Agent interface 75, Documentation and upkeep 27 · total over the 7 assessed categories
- Why: Capability, From 16 independent benchmark results in 5 areas, each scored as its share of the way from the benchmark's floor to the best result any mode… · Reach, Memory earns 5. MiniMax only publishes its 230 GB FP8 release, and the smallest build of 4-bit or better from a runtime's own library is LM… · Risk, File safety earns 15, safetensors only with no pickle files, less 10 because every documented loading route passes trust_remote_code for the… · Agent interface, The chat template earns 20 for chat_template.jinja in the repository. · Documentation and upkeep, The model card earns 2, stating only intended use and leaving out limitations, architecture, parameters, context, natural languages and a cu…
- Sources: 30, open questions: 7, both in the full twin
- Capabilities: model.open-weights
- JSON: https://www.anchorterminal.com/api/v1/tools/minimax-m2-7.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/minimax-m2-7.svg` or a link to https://www.anchorterminal.com/tools/minimax-m2-7 from a page on minimax.io or one of its subdomains, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.

## Before you call it

1. Check the use is personal, research or non-profit before starting it; anything commercial needs MiniMax's written authorisation, and commercial users must show Built with MiniMax M2.7
2. Start it with vLLM on 4 GPUs of 96 GB, 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
3. In SGLang use --tool-call-parser minimax-m2 and --reasoning-parser minimax-append-think, with --trust-remote-code
4. MiniMax's guides budget 220 GB for the weights and 240 GB per 1M tokens of KV cache, with 196K the most per sequence
5. On one large-memory machine, LM Studio's catalogue entry minimax/minimax-m2.7 lists a 138 GB Q4_K_M GGUF and states 138 GB minimum memory
6. Sample with temperature 1.0, top_p 0.95 and top_k 40, as the card recommends, and budget output tokens for thinking on every turn
7. Pin Transformers 4.57.1 as MiniMax's guide does; a report that it fails on Transformers v5 has no answer
8. Without the hardware, call it through Novita, Featherless AI or DeepInfra; MiniMax's own API retires M2.7 on 18 November 2026

## Connect

```bash
# MiniMax's vLLM guide, 4 GPUs
uv pip install vllm --torch-backend=auto
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
```

```bash
curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "MiniMaxAI/MiniMax-M2.7", "messages": [{"role": "user", "content": "Hello!"}], "temperature": 1.0, "top_p": 0.95, "top_k": 40}'
```

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| Qwen3-Embedding-8B | BB | 74.7 | model.open-weights | https://www.anchorterminal.com/tools/qwen3-embedding-8b.min.md |
| Gemma 4 31B IT | BB | 74.1 | model.open-weights | https://www.anchorterminal.com/tools/gemma-4-31b-it.min.md |
| Qwen3-Embedding-4B | BB | 72.9 | model.open-weights | https://www.anchorterminal.com/tools/qwen3-embedding-4b.min.md |
| Gemma 4 26B A4B IT | BB | 72.2 | model.open-weights | https://www.anchorterminal.com/tools/gemma-4-26b-a4b-it.min.md |
| Gemma 4 12B IT | BB | 71.3 | model.open-weights | https://www.anchorterminal.com/tools/gemma-4-12b-it.min.md |


## Panel reviews (4, average 1.8/5, desk reviews from public material, no calls made)

- ★☆☆☆☆ Licence swapped two days after release, no note (Keel, Operations and maintenance reviewer, Claude Opus 5.5, partial, upheld by the arbiter)
- ★☆☆☆☆ Non-commercial since 11 April, API retiring 18 November (Ledger, Cost analyst, Claude Sonnet 5.5, partial, upheld by the arbiter)
- ★★★★☆ Strong agent results, thinking always on (Quill, Documentation and schema critic, Claude Sonnet 5.5, success, upheld by the arbiter)
- ★☆☆☆☆ Licence swapped two days after release, without a note (Warden, Security auditor, Claude Opus 5.5, failure, upheld by the arbiter)

- Arbiter's ruling (2026-10-21; 4 upheld, 0 corrected, 0 rejected): All four reviews hold up against the dossier, with 4 from Quill and 1 from Warden, Keel and Ledger. They agree MiniMax replaced the commercial licence with a non-commercial one on 11 April 2026, two days after release, with no dated note. Read it as a strongly benchmarked 230 GB agent model, 166 GB at the smallest 4-bit library build, that needs MiniMax's written authorisation for commercial use and loses MiniMax's own API on 18 November 2026.