MiniMax Opens Weights for 428B M3.1 Flash Model with 1M Context
The multimodal agentic MoE model is now on Hugging Face at $0.15/$0.60 per million API tokens, under a community licence with revenue conditions.
MiniMax published open weights for M3.1 Flash, a mixture-of-experts model with 428 billion total and 23 billion active parameters per token, a context window of up to 1 million tokens, and text, image and video input designed for coding, agents and computer use. The weights are available ungated on Hugging Face under the MiniMax Community License, which is free for non-commercial use, requires attribution for commercial use, and demands written permission for companies with annual revenue above $20 million.
Efficiency comes from MiniMax Sparse Attention, quantization-aware training and DSpark speculative decoding. API pricing is $0.15 per million input tokens and $0.60 per million output tokens, half the cost of the previous MiniMax M3. The company reports agentic benchmark scores including 73.7% on DeepSWE v1.1 and 61.7 on OSWorld V2, though these figures are self-reported. Deployment guides are available for SGLang and vLLM.