✦
Summary
Muse Ember is Meta's smallest and cheapest Muse-family model, released on November 9, 2026 for the high-volume repetitive work AI agents do all day: tool calls, coding subtasks, routing, data extraction and long-document reading. It is a multimodal model that natively understands text, image, audio and video, with a 1-million-token context window and up to 512,000 output tokens per response. It is API-only with no open weights, and acts as a cheap subagent or worker model alongside stronger models.
Key Features
1-million-token context window
Up to 512,000 output tokens per response
Multimodal input: text, image, audio and video
Built for repeated agent subtasks: tool calls, coding subtasks, routing, data extraction and long-document reading
Smallest and cheapest Muse model, priced from $0.10 per 1M input tokens up to 200K tokens
Available on the Meta Model API, Muse Code, Google Cloud, Oracle Cloud and OpenRouter
Supports the Responses API
Scores 36 on the Artificial Analysis Intelligence Index, level with Claude Haiku 5.5
About Muse Ember
Pricing is $0.10 per 1M input tokens and $0.50 per 1M output tokens for prompts up to 200K tokens, rising to $0.30 input and $1.50 output above 200K tokens, with cached input at $0.004 per 1M tokens in the lower tier. It scores 36 on the Artificial Analysis Intelligence Index, level with Claude Haiku 5.5, and supports the Responses API. Access is through the Meta Model API as muse-ember-1.0, inside Muse Code, and via Google Cloud, Oracle Cloud and OpenRouter.
Muse Ember is included in Muse Code subscription plans, which start at $5 per month, and is aimed at developers running agents that make thousands of small model calls a day, where per-call cost matters more than maximum reasoning depth. Because it is the smallest Muse model, it trades some capability for speed and price, and is designed to sit next to a larger model that handles the hard steps while Ember handles the routine ones.