Meta introduced Muse Ember in Muse Code and on the Meta Model API. The developer announcement describes a model for recurring coding subtasks, tool calls and document processing. Meta suggests using it on bounded jobs and pairing it with a larger model when the work needs more planning.
Which inputs can Muse Ember handle?
Muse Ember accepts text, image, audio and video inputs. Meta gives examples such as extracting information from scanned documents, reading a screenshot alongside a bug report or summarizing action items from audio. Those multimodal inputs let a workflow pass the original materials to the model. The announcement describes reading and analyzing them; it does not say Ember generates those media types.
Where is it available, and how is it priced?
Meta says Muse Ember is available across all Muse Code subscription plans. Its API announcement lists $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 200,000 input tokens. For prompts above that threshold, the listed rates are $0.30 and $1.50 per million input and output tokens, respectively. The lower listed rate therefore depends on prompt size.
Meta also describes a path for projects using an OpenAI-compatible client: point it at the Meta Model API, use a Meta key and select the Ember model. Teams should confirm which request options they rely on before assuming every part of their existing integration behaves the same.
How does Meta suggest dividing work?
Meta shows a parallel code migration and a batch photo organization workflow. It suggests using Muse Wonder or Muse Spark for deeper planning while Ember handles bounded steps. That division gives teams a measurable unit of work to test, such as reviewing one module or extracting fields from one document, before trusting the orchestration across a larger batch.
