Meta's small multimodal model for high-volume agent work
Grounded in available product and source data
Muse Ember is Meta's smallest multimodal model, built for the work agents repeat all day. It handles coding subtasks, tool calls, routing, extraction and long document summarization, and accepts text, image, audio and video inputs.
In practice it runs up to 30% faster at the tail than a comparable model in one early test, with fewer output tokens. Meta pairs it with Muse Wonder and Muse Spark so a team can match task complexity to model capability.