Muse Spark Open

NEW

Summary

Muse Spark Open is Meta's open-weight multimodal model for long-horizon coding and agentic reasoning. Its mixture-of-experts architecture has 816B total and 42B active parameters with a one-million-token context window.

Key Features

Natively multimodal inputs.
One-million-token context.
Agentic coding.
Multi-step reasoning.
Public model weights.
MLX support.
vLLM and SGLang support.
PyTorch customization.

About Muse Spark Open


Developers can customize and host the model across major inference engines, including supported multi-machine Apple Silicon deployments. Core capabilities include Natively multimodal inputs, One-million-token context, Agentic coding, and Multi-step reasoning.


Public code or model weights are linked from the project, with setup instructions for running or adapting the system. Additional characteristics include Public model weights, MLX support, vLLM and SGLang support, and PyTorch customization.

Get more likes & reach the top of search results by adding this button on your site!

Embed button preview - Light theme
Embed button preview - Dark theme
TurboType Banner

Subscribe to the AI Search Newsletter

Get top updates in AI to your inbox every weekend. It's free!