THE DAILY EDITION · 15 OCTOBER 2026
Meta unveils Muse Spark Open, an open-weight multimodal model
Why this story?
Meta’s release combines text, image, audio and video in a single open-weight model with a reported one-million-token context window, making the scale of the system notable.
Meta has released Muse Spark Open, an open-weight model that accepts text, images, audio and video and generates text from those inputs, according to material published on Hugging Face. The model is described as a mixture of experts with about 816 billion parameters, around 42 billion active per token, and a context window of up to one million tokens. The same material says it gives developers a common interface for asking questions about different kinds of input.
Muse Spark Open follows Muse Glimmer and introduces a mixture-of-experts architecture at a much larger scale. The Hugging Face post also describes Muse Spark Open as the most capable US open-weights model to date.
The post includes examples of sending requests to the model through an OpenAI-compatible API and through the same Hub model ID, showing how the model is intended to be used by developers.