Muse Ember

Muse Ember is Meta's smallest and most efficient model in the Muse family, built for high-volume, repeated work that AI agents do all day. It handles tasks such as making tool calls, running focused coding subtasks, routing and classifying requests, and extracting information from many files or records at once. For example, a team can have it caption, rename, and sort thousands of photos into albums, or read screenshots, scanned documents, voice memos, and video directly without a separate model for each.

It suits teams that need to run the same kind of small task many times and want each one handled quickly and at low cost. Muse Ember is meant to work alongside larger Muse models such as Muse Wonder or Muse Spark, which do the planning and harder reasoning while Muse Ember carries out the high-volume execution. It is a proprietary model, so its weights are not publicly available.