Muse Spark Open

Muse Spark Open is available today and brings leading agentic capabilities to your desktop for the first time on Apple M5 Ultra Mac Studios. Today, we’re releasing the weights for Muse Spark Open, the most capable US open-weight model to date. Muse Spark Open is a natively multimodal model with 816B total parameters (42B active) and a 1M-token context window. Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning.

Muse Spark Open 于今日发布，首次在 Apple M5 Ultra Mac Studio 上为您的桌面端带来了领先的智能体能力。今天，我们发布了 Muse Spark Open 的权重，这是迄今为止美国最强大的开源权重模型。Muse Spark Open 是一个原生多模态模型，拥有 8160 亿总参数（420 亿激活参数）和 100 万 token 的上下文窗口。它专为编程和智能体工作而构建，擅长处理长周期任务、复杂的软件工程以及多步推理。

We believe the future is for everyone, and that includes the freedom to run advanced models wherever, including at home. Muse Spark Open is the first model in its class to run on your desktop using two Apple M5 Ultra Mac Studios with the latest MLX framework. Muse Spark Open is available today through API platforms at Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter and Together AI. Model weights are available for download on Amazon SageMaker AI, Google Cloud Model Garden, Hugging Face, and Microsoft Foundry Models. It runs on leading inference engines like SGLang, vLLM, TensorRT-LLM, Hugging Face Transformers, Apple MLX, llama.cpp, and Unsloth with optimizations for AMD, Apple, Intel, and NVIDIA silicon. Developers can easily further customize the model using the PyTorch framework.

我们相信未来属于每一个人，这包括在任何地方（包括家中）自由运行先进模型的权利。Muse Spark Open 是同类模型中首个能够通过两台搭载最新 MLX 框架的 Apple M5 Ultra Mac Studio 在桌面端运行的模型。Muse Spark Open 今日起可通过 Baseten、Fireworks AI、Groq、Modal、Nebius、OpenRouter 和 Together AI 等 API 平台获取。模型权重可在 Amazon SageMaker AI、Google Cloud Model Garden、Hugging Face 和 Microsoft Foundry Models 上下载。它可以在 SGLang、vLLM、TensorRT-LLM、Hugging Face Transformers、Apple MLX、llama.cpp 和 Unsloth 等主流推理引擎上运行，并针对 AMD、Apple、Intel 和 NVIDIA 芯片进行了优化。开发者可以使用 PyTorch 框架轻松地进一步定制该模型。

Key Benchmarks For more details about our evaluations, see the evaluation methodology report. Optimized for Apple Silicon Leveraging the latest version of MLX, Muse Spark Open was optimized to run on Mac Studio with M5 Ultra, and together they’re the best way to run a powerful model on your own desktop. Apple’s unified memory architecture lets the M5 Ultra’s integrated GPU use up to 512 GB per Mac Studio, so two machines can hold an entire model. And the chip’s 1.2 TB/s of memory bandwidth is fast enough to read the small slice each token needs. The result is two Mac Studios behave like one supercomputer with a terabyte of unified memory.

关键基准测试：有关我们评估的更多详细信息，请参阅评估方法报告。针对 Apple 芯片优化：利用最新版本的 MLX，Muse Spark Open 经过优化可在搭载 M5 Ultra 的 Mac Studio 上运行，它们共同构成了在个人桌面端运行强大模型的最佳方式。Apple 的统一内存架构允许 M5 Ultra 的集成 GPU 为每台 Mac Studio 使用高达 512 GB 的内存，因此两台机器可以容纳整个模型。此外，该芯片 1.2 TB/s 的内存带宽足以读取每个 token 所需的小片段。其结果是两台 Mac Studio 表现得就像一台拥有 1 TB 统一内存的超级计算机。

We tested a range of agentic workflows running entirely on-device, using Muse Code powered by Muse Spark Open, including shipping real iOS changes to Ice Cubes (an open-source Mastodon client) — bug fixes, new features, full build — and driving DaVinci Resolve end to end, with the model reviewing footage frame by frame to find the strongest moments and crop them vertically. Available Now We’re sharing this work because we want everyone to have access to the best models. The open-weight version of Muse Spark ships with custom, auto-tunable Helion GPU kernels in SGLang, delivering 1.03-3.95× kernel-level speedups on NVIDIA and AMD GPUs, and a 3% end-to-end speedup on AMD MI350.

我们测试了一系列完全在设备上运行的智能体工作流，这些工作流使用了由 Muse Spark Open 驱动的 Muse Code，包括向 Ice Cubes（一款开源 Mastodon 客户端）发布真实的 iOS 更改——包括错误修复、新功能、完整构建——以及端到端驱动 DaVinci Resolve，模型逐帧审查素材以找到最精彩的片段并进行垂直裁剪。现已发布：我们分享这项工作是因为我们希望每个人都能获得最好的模型。Muse Spark 的开源权重版本随附 SGLang 中自定义的、可自动调优的 Helion GPU 内核，在 NVIDIA 和 AMD GPU 上实现了 1.03-3.95 倍的内核级加速，并在 AMD MI350 上实现了 3% 的端到端加速。

Safety and Preparedness Releasing open-weight models is part of how we bring the benefits of personal superintelligence to everyone, and part of how safety work advances across the field. Open weights also carry their own considerations, and Muse Spark Open has additional safety post-training for that reason. The results demonstrate strong safeguards across risk categories measured. We evaluated the model in line with our Meta Superintelligence Scaling Framework, covering the most severe risks, and it met the thresholds our framework requires for release. Detailed results are in the Muse Spark Open model card, with a full evaluation report available soon.

安全与准备：发布开源权重模型是我们为每个人带来个人超级智能益处的方式之一，也是推动整个领域安全工作进步的一部分。开源权重也有其自身的考量，因此 Muse Spark Open 进行了额外的安全后训练。结果证明了在所测量的风险类别中具有强大的保障措施。我们根据 Meta 超级智能扩展框架对模型进行了评估，涵盖了最严重的风险，它达到了我们框架要求的发布阈值。详细结果见 Muse Spark Open 模型卡，完整的评估报告即将发布。

License We’re releasing the weights under the Muse Model License with limited commercial terms, which allow many developers to use the model for free. Get Started You can customize Muse Spark Open for your use case by leveraging PyTorch’s TorchTitan to further tune the model. To run it on Apple silicon, use MLX, which pulls the weights directly from Hugging Face. Muse Spark Open is available now through Hugging Face: https://huggingface.co/collections/meta-models/muse-spark-open MLX Adapter for running Muse Spark Open on Mac Studio. Hosted Inference Partners: Together AI, Fireworks, Baseten, Modal, Groq, Nebius, and OpenRouter. Built or published something with Muse Spark Open? Share with us on the Hugging Face Forums — we’d love to see what you create.

许可：我们根据 Muse 模型许可发布权重，该许可包含有限的商业条款，允许许多开发者免费使用该模型。开始使用：您可以利用 PyTorch 的 TorchTitan 对模型进行进一步微调，从而为您的用例定制 Muse Spark Open。要在 Apple 芯片上运行它，请使用 MLX，它会直接从 Hugging Face 拉取权重。Muse Spark Open 现已通过 Hugging Face 提供：https://huggingface.co/collections/meta-models/muse-spark-open，以及用于在 Mac Studio 上运行 Muse Spark Open 的 MLX 适配器。托管推理合作伙伴：Together AI、Fireworks、Baseten、Modal、Groq、Nebius 和 OpenRouter。使用 Muse Spark Open 构建或发布了什么作品？请在 Hugging Face 论坛上与我们分享——我们很期待看到您的创作。