Welcome Muse Spark Open: open-weight frontier multimodal mixture-of-experts model from Meta

Meta has released Muse Spark Open, a frontier open-weight model that processes text, images, audio, and video inputs. It utilizes a massive mixture-of-experts architecture with 816 billion parameters, but only activates around 42 billion per token for efficient computation. The model is specifically optimized for agentic coding, enabling it to plan, act, and self-correct on tasks like multi-file refactoring and long debugging sessions. With a context window of up to one million tokens, it can analyze entire codebases or watch screen recordings to complete complex developer workflows.