Muse Ember is out, Meta's small model for agent work

Meta released Muse Ember, its smallest model, on the Meta Model API and Muse Code. The API name is muse-ember-1.0.

Ember is built for the work agents repeat all day: tool calls, coding subtasks, routing, extraction, and long document parsing. The context window is 1M tokens (a token is a small unit of text). Output goes up to 512K tokens. The price is $0.10 per million input tokens and $0.50 per million output. Cached input is $0.004.

The point is routing. When an agent makes the same small call hundreds of times a session, the cheap model doing it well beats the big model doing it expensively. Meta says Ember is the first model trained using Muse Wonder Full, and the rule is simple: Wonder plans, Ember handles the volume.

Here is how I am testing it. A Wonder Full lead lists my PDFs and writes a small script that sends each one to Ember through the Model API. Ember workers mark each file keep, skip, or duplicate, and write Markdown for the keepers. The lead checks the results. I keep each Markdown file in Hindsight, which stores the facts it extracts as memories.

Primary source: Meta research blog. Docs: Choose a model, Custom agents and Organize photos with subagents. Coverage: TechCrunch.