MiniMax发布了其M3模型，该模型采用了新颖的稀疏注意力（MSA）架构，支持100万token的上下文窗口和原生多模态能力。这种新架构显著降低了长上下文的计算成本，使得M3比前几代产品速度更快。该模型在编码和智能体任务方面也表现出色，在SWE-Bench Pro和Terminal-Bench等基准测试中超越了多个领先模型。 AI
影响 在编码基准测试中设定了新的SOTA（State-of-the-Art），并提供了前所未有的上下文长度，可能改变行业在模型效率和能力方面的标准。
排序理由 前沿实验室模型发布，附带系统卡和新架构。
AI 生成摘要 · Google Gemini · 来自 8 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
MiniMax M3 arrives with MiniMax Sparse Attention (MSA), 15.6x faster decoding at 1M tokens. We're partnering with @MiniMax_AI to power the inference behind this week's launch. Head to https://t.co/zwLs8Pj7I6 to take it for a spin. Once the model weights are released, M3 will be …
Speakers: Pengyu Zhao, Head of Research at MiniMax Haohai Sun, Research Scientist at MiniMax Ce Zhang, Founder/CTO at Together Yineng Zhang, Senior Director at Together Dan Fu, VP of Kernels at Together Hosted by Zain Hasan, Staff AI Engineer at Together
We'll get into @MiniMax_AI M3's model performance, the MSA architecture and what it means for long context, and how Together is optimizing inference and KV-cache for this new architecture. Set your reminders!
Speakers: Pengyu Zhao, Head of Research at MiniMax Haohai Sun, Research Scientist at MiniMax Ce Zhang, Founder/CTO at Together Dan Fu, VP of Kernels at Together Hosted by Zain Hasan, Staff AI Engineer at Together
We'll get into @MiniMax_AI M3's model performance, the MSA architecture and what it means for long context, and how Together is optimizing inference and KV-cache for this new architecture. Set your reminders.
MiniMax M3 is live and Together AI is powering its inference 🚀 Tomorrow at 6pm PT we're going live on X Spaces with the teams behind the model and the infrastructure to give you a deep dive. https://t.co/wPayfOWmNg
How Together served MiniMax-M3 efficiently with KV-block-major sparse attention, paged MSA decode, optimized index scoring, and a Rust-based multimodal gateway.
<p>MiniMax M3 introduces MiniMax Sparse Attention, a 1M-token context window, and native image, video, and computer use support.</p> <p>The post <a href="https://www.marktechpost.com/2026/06/01/minimax-releases-minimax-m3-with-msa-architecture-supporting-1m-token-context-native-m…