Meta released Muse Ember as a smaller model for repeated operations within workflows carried out by artificial intelligence agents. Its intended tasks include tool use, bounded coding work and information extraction. The model can work alongside the larger Muse Wonder, separating demanding planning from routine execution within Meta’s developer products.
Meta released Muse Ember on November 9, adding a smaller model to its Muse artificial intelligence family. The company presents Ember as an execution component for work carried out by software agents. Its intended tasks include calling tools, routing requests, extracting information and carrying out bounded coding subtasks within a larger workflow.[1], [2]
Ember can work alongside Muse Wonder, the larger model in Meta’s product family. A customer-support example divides work between a larger model that plans and a smaller model that performs repeated operations. This allocates different parts of one task to different models. Long-document work is another intended use, alongside information extraction and software execution.[1]
Ember accepts text, images, audio and video as inputs. Meta is adding it to the existing Muse tools used by application developers. The launch follows Spark in April, Glimmer in August and Wonder in October; Muse Agent arrived in September. Ember’s stated role is repeated execution within these developer workflows, alongside models handling more demanding reasoning and planning.[1]
