Meta Superintelligence Labs introduced Muse Ember on Monday, Nov. 9. It is the smallest and cheapest Muse model so far. Meta built it for the repetitive steps inside AI agents rather than for hard reasoning.
What it does
Meta says Ember handles coding subtasks, tool calls, routing, and document extraction. It accepts text, image, audio, and video inputs. It is available in Muse Code, Meta’s coding agent, and through the Meta Model API. TechCrunch notes that Ember is the fourth model under the Muse brand, after Spark in April, Glimmer in August, and Wonder in October.
Price
Meta lists $0.10 per million input tokens and $0.50 per million output tokens for prompts of up to 200,000 input tokens. Longer prompts cost $0.30 and $1.50.
How Meta wants it used
The design is a two-tier setup. TechCrunch quotes Meta: “Muse Wonder plans, Muse Ember handles the volume.” A larger model breaks a job into pieces, and many small Ember agents work on those pieces in parallel.
As a demo, Meta ran 16 Ember agents to port 16 syntax-highlighting lexers from Python to Rust. Meta says the agents finished within about 6 minutes of each other.
The numbers, and a caveat
Meta says Ember scores 36 on the Artificial Analysis Intelligence Index. At maximum reasoning effort, Meta estimates the cost at about $0.13 per index task. MarkTechPost reports that the 36 ties Claude Haiku 5.5.
On AutomationBench-AA, a business-workflow test, Meta lists 37.8 for Ember against 13.9 for Haiku 5.5 and 20.2 for GPT-6 Luna. These are Meta’s published results, and Meta chose the models it compared against. Independent testing will show whether the cost advantage holds on other workloads.
