开源模型首次在编码和智能体任务上追平 GPT-5.5 和 Opus，成本却低一个数量级。做 AI 应用开发或自建模型的团队，值得关注权重发布后直接试用。
MiniMax 发布开源权重模型 M3，在 SWE-Bench Pro 上达到 59.0%，与 GPT-5.5 持平，并在多项编码基准上与 Opus 4.7 不相上下。M3 还支持 1M 上下文、原生多模态，使用成本仅为 GPT 和 Opus 的十分之一。权重和技术报告将在约 10 天后在 Hugging Face 开放。这标志着开源模型首次在多个前沿能力上追平闭源顶级模型，对 AI 开发者和企业用户意义重大。
Open source is going to win We already have an open-weights model competitive with GPT-5.5 and Opus...
Open source is going to win We already have an open-weights model competitive with GPT-5.5 and Opus 4.7 🤯 MiniMax M3 is genuinely impressive: - On par w/ GPT-5.5 on SWE Bench Pro - Neck and neck w/ Opus 4.7 on many - Better than Gemini 3.1 Pro for coding But MORE IMPORTANT: M3 is 10x cheaper to use than both GPT and Opus!! And the weights will be available on Hugging Face next week. We're so back. MiniMax (official) @MiniMax_AI Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities - Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas - MiniMax Sparse Attention scales context to 1M - Natively Multimodal from Step Zero API: platform.minimax.io Token Plan: platform.minimax.io/subscribe/toke… 🚀New! MiniMax Code: code.minimax.io l Weights & Tech Report in ~10 Days 🔗 View Quoted Tweet 💬 15 🔄 3 ❤️ 58 👀 6465 📊 16 ⚡