Meta releases Muse Ember, a small model for repeated agent tasks

Meta's post dated 9 November 2026 introduces Muse Ember. It describes the model as its smallest and most efficient, for tool calls, coding subtasks, routing, extraction and long document parsing. It says Muse Ember is available the same day on Muse Code and the Meta Model API as muse-ember-1.0. It prices the model at $0.10 per million input tokens, $0.004 per million cached input tokens and $0.50 per million output tokens, with a 1M-token context window and up to 512K output tokens. It says the model supports the Responses API.

The post says Muse Ember is the first model trained using Muse Wonder Full. It says that, before release, Meta evaluated it under the Meta Superintelligence Scaling Framework, including chemical and biological, cybersecurity and loss-of-control risks, and that across the categories assessed it remained below the framework's high-risk thresholds. The post does not say the weights are being released.

What it means for business: this is a priced API model for high-volume steps, at the rates above. Meta's own evaluation says it stayed below that framework's high-risk thresholds. The post does not publish weights. This describes the announcement. It is not advice on sending business data to the API.