Muse Ember, a new model
Muse Ember 1.0, from Meta, is a new model in the chat’s model selector, next to DeepSeek V4.1 Flash and GLM 5.3 Flash. It reads images and reasons before answering: High by default, and you can pick Low, Medium or Max in the Effort row. Each answer uses less of your usage than with the other two.
Unlike the other two, Muse Ember is not an open model: Meta runs it on its own API, in the US, so Meta receives what you send to it. It has a single provider: if Meta is down, the answer ends with an error and Retry, and you can switch to another model in the selector.
Admins choose which models their organization can use, and which one is the default, in Organization → Models. More about the selector and effort in Writing a request.