Key Info
Meta released Muse Spark Open, described as the most capable US open-weight model to date, with 816B total parameters, 42B active parameters, a 1M-token context window, and native multimodality across text, image, video, and audio.
Highlights
- Built for coding and agentic work, targeting long-horizon tasks, complex software engineering, and multi-step reasoning.
- Optimized to run on two Apple M5 Ultra Mac Studios with the latest MLX framework, and supported by SGLang, vLLM, TensorRT-LLM, Hugging Face Transformers, llama.cpp, and Unsloth.
- Available via API platforms including Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter, and Together AI.
- Weights can be downloaded on Amazon SageMaker AI, Google Cloud Model Garden, Hugging Face, and Microsoft Foundry Models under the Muse Model License with limited commercial terms.
- Developers can customize and fine-tune with PyTorch, including TorchTitan.