Tier 2 Restricted open-weight
What the publisher claims: open-source, open-weight
The one-liner: Frontier-scale MoE family whose openness tightens release over release, from modified MIT to outright non-commercial.
The six questions
weights downloadable from Hugging Face and ModelScope for M2, M2.7 and M3
M2.7 is non-commercial by license text; M3 imposes branding and notice/authorization duties on commercial use
serves under Transformers, vLLM and SGLang; also on NVIDIA NIM
architecture, parameters, benchmarks and blog posts published; training corpus and recipe withheld
cannot rebuild a substantially equivalent model from published material
MiniMax publishes weights, modeling code, and eval results on Hugging Face with issue trackers open but no development roadmap and no community discussion, so every model decision is made inside the company.
License
Modified MIT (M2); non-commercial MIT-style (M2.7); MiniMax Community License (M3) - verify per release — license source ↗
- M2 adds a branding clause only for products over 100M MAU or 30M USD ARR
- M2.7 permits non-commercial use only and requires prior written authorization for any commercial use
- M3 grants non-commercial use and requires Built with MiniMax M3 display plus a notice or authorization above 20M USD yearly revenue
What is actually released
- weights (M2, M2.7, M3)
- custom modeling code
- tokenizer
- eval results
- blog posts and tech reports; no training data or training code
Why it matters
Shows that open weights can be shipped under a source-available reading of the word open, with the same vendor marking M2 open-source in a README while its own M2.7 LICENSE restricts commercial use.
Sources
- https://huggingface.co/MiniMaxAI/MiniMax-M2.7
- https://huggingface.co/MiniMaxAI/MiniMax-M2/raw/main/README.md
- https://raw.githubusercontent.com/MiniMax-AI/MiniMax-M2/main/LICENSE
- https://huggingface.co/MiniMaxAI/MiniMax-M2.7/raw/main/LICENSE
- https://huggingface.co/MiniMaxAI/MiniMax-M3/raw/main/LICENSE
- https://huggingface.co/api/models/MiniMaxAI/MiniMax-M2.7
- https://huggingface.co/api/models/MiniMaxAI/MiniMax-M3
Derived from the openness dataset behind But Is It Open? — snapshot 2026-11-07 ↗. Scores are qualitative readings of the cited evidence, not verdicts; tier measures openness, not intelligence.