MiniMax M3 open-source MoE model beats GPT-5.5 and Claude Opus 4.7 on SWE-bench
Key Questions
What are the key specifications of the MiniMax M3 model?
MiniMax M3 is a 229.9B parameter open-source MoE model with a 1M context window and an innovative sparse attention mechanism called MSA. It also emphasizes multimodal capabilities as part of its design.
How does MiniMax M3 compare to other leading models on benchmarks?
The model outperforms Claude Opus 4.7 and GPT-5.5 on SWE-bench, and it also surpasses GPT-5.5 and Gemini 3.1 Pro on key benchmarks. This performance is achieved at only 5-10% of the typical cost.
When will MiniMax M3 weights be released and what is the pricing strategy?
Weights for the open-source model are scheduled for release in 10 days. Aggressive pricing has been announced to support broader accessibility.
MiniMax M3 is a 229.9B parameter open-source MoE model with 1M context and innovative sparse attention (MSA). Benchmarks show it beating Claude Opus 4.7 and GPT-5.5 on SWE-bench. Weights to be released in 10 days. Aggressive pricing announced.