MiniMax 发布原生多模态模型 MiniMax-M3，支持 1M 上下文
MiniMaxAI/MiniMax-M3
MiniMax 发布原生多模态模型 MiniMax-M3，参数量约 428B、激活参数约 23B，支持 1M 上下文。该模型从训练第一步起就进行混合模态训练，并引入 MiniMax Sparse Attention（MSA），在 1M 上下文下相比 M2 实现 9 倍 prefill 和 15 倍 decode 加速，单 token 计算量降至 1/20。
原文给出参数量、稀疏注意力和长上下文加速数据，读者可据此判断百万 token 场景的部署成本变化。
MiniMax-M3 is a native multimodal model with 1M context. It has ~428B parameters and ~23B activated parameters.
Highlights:
- Native Multimodality: M3 undergoes mixed-modality training from the very first step, enabling deeper semantic fusion across text, image, and video.
- Context Scaling via Sparse Attention: M3 introduces MiniMax Sparse Attention (MSA) to improve long context efficiency. M3 delivers 9× prefill and 15× decode speedups compared to M2 at 1M context, reducing per-token compute to 1/20.
- Coding & Cowork Capability: M3 achieves frontier-level performance across long-horizon agentic benchmarks, excelling in both coding and cowork.
MiniMax Sparse Attention (MSA)
M3 is powered by MiniMax Sparse Attention (MSA), a high-performance sparse attention operator designed for million-token contexts. Compared with GQA, MSA dramatically reduces the attention compute and memory footprint while preserving model quality.
📄 Read the technical report: arXiv:2606.13392 · HF Mirror Papers
How to Use
M3 supports three reasoning modes through the thinking parameter:
enabled— Reasoning is always enabled.adaptive— M3 automatically determines when additional reasoning is beneficial.disabled— Reasoning is disabled to minimize latency and maximize throughput.
Local Deployment
Download the model:
hf download MiniMaxAI/MiniMax-M3 --local-dir MiniMax-M3
We recommend the following inference frameworks to serve the model:
SGLang - see SGLang cookbook.
vLLM - see vLLM recipes.
Transformers - see Transformers docs.
Inference Parameters
We recommend the following parameters for best performance: temperature=1.0, top_p=0.95.
Contact Us
Contact us at [email protected].
来源：MiniMax 模型发布 · hf-mirror.com