Meta introduced Muse Ember on November 9 as its smallest and most efficient model, targeting the many routine steps that agents perform between larger planning tasks. The company highlights tool use, code subtasks, routing, information extraction and long-document processing rather than positioning Ember as its largest frontier reasoning model.
Meta says Ember supports text, image, audio and video understanding and a context window reaching one million tokens. Its proposed standard API prices start at $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 200,000 tokens, with different prices for longer prompts.
Meta compares cost per task against Anthropic's Claude Haiku 5.5 and OpenAI's GPT-6 Luna, citing Artificial Analysis scores alongside its own evaluations. These comparisons are not universal evidence of superiority: harnesses, workloads, reasoning settings and benchmark selection can change rankings, and Meta notes some figures are estimates.
The commercial significance is an increasingly differentiated market for small, economical models within multi-agent systems. Organizations can reserve expensive models for planning or hard reasoning and use lighter models for high-volume execution, but actual savings depend on reliability and orchestration overhead.
Meta says Ember is available through its model API and supported cloud distribution channels. That is an announced product capability, not proof that all developers have adopted it or achieved the company's claimed economics.