Meta 发布开源权重模型 Muse Spark Open，816B 总参数、1M 上下文
To help make advanced models more accessible, Muse Spark Open is optimized to run on two Apple M5 Ultra Mac Studios with the latest MLX framework.
Meta 发布开源权重模型 Muse Spark Open，面向编码与智能体任务，总参数 816B、激活参数 42B，上下文窗口 1M token，原生支持文本、图像、视频和音频。该模型已针对双 Apple M5 Ultra Mac Studio 与最新 MLX 框架优化，同时支持 SGLang、vLLM、TensorRT-LLM、Hugging Face Transformers、llama.cpp 和 Unsloth 等推理引擎，并针对 AMD、Apple、Intel 和 NVIDIA 芯片优化。模型今日起在 Baseten、Fireworks AI、Groq、Modal、Nebius、OpenRouter 和 Together AI 等 API 平台上线，权重可在 Amazon SageMaker AI、Google Cloud Model Garden、Hugging Face 和 Microsoft Foundry Models 下载，采用 Muse Model License 发布，开发者可用 PyTorch 及 TorchTitan 进一步微调。
正文
当前显示原文，中文全文译文尚未就绪。
To help make advanced models more accessible, Muse Spark Open is optimized to run on two Apple M5 Ultra Mac Studios with the latest MLX framework.
It also runs on leading inference engines like SGLang, vLLM, TensorRT-LLM, Hugging Face Transformers, llama.cpp, and Unsloth with optimizations for AMD, Apple, Intel, and NVIDIA silicon.
Developers can further customize the model with PyTorch, including TorchTitan for fine-tuning.