Meta 在 NVIDIA GB300 NVL72 上开源 Muse Spark Open 编码模型
Scale Agentic Coding Workflows with Meta’s Muse Spark Open on NVIDIA GB300 NVL72
Meta 发布编码优化模型 Muse Spark Open，采用 816B 参数的 MoE 架构，激活参数 42B，上下文窗口 1M token。该模型专门基于 NVIDIA GB300 NVL72 Blackwell Ultra 训练，面向智能体编码工作流，具备更高的首次尝试准确率、可靠的工具调用和跨仓库的深度代码库理解能力。
Meta 开源 Muse Spark Open 的架构规格与 NVIDIA GB300 NVL72 上的智能体编码表现，可作为模型选型参考。
Meta launches Muse Spark Open, a coding-optimized 816B mixture-of-experts (MoE) model with 42B active parameters and a 1M token context window. Trained on...
Meta launches Muse Spark Open, a coding-optimized 816B mixture-of-experts (MoE) model with 42B active parameters and a 1M token context window. Trained on NVIDIA GB300 NVL72 Blackwell Ultra specifically for agentic coding workflows, it delivers higher first attempt accuracy, reliable tool calling, and deep codebase understanding across repositories. The 1M token context window holds an entire…
来源：NVIDIA Developer Blog · developer.nvidia.com