What changed
Today, Meta introduced Muse Ember, their smallest and most efficient model designed for complex reasoning execution and the tasks agents repeat daily, such as tool calls, coding subtasks, routing, extraction, and long document parsing. Muse Ember is positioned as a cost-effective alternative to models like Claude Haiku 5.5 and GPT-6 Luna, delivering more intelligence per dollar.
What this means in practice
Muse Ember is available today on the Meta Model API, with pricing at $0.10 per million input tokens and $0.50 per million output tokens.
It is built for workflows involving many small, structured steps, such as reading files, running tests, and making targeted fixes.
The model has a 1M-token context window and can handle up to 512K output tokens, allowing it to process a full codebase or document set in a single call.
It also supports multimodal tasks, understanding text, images, audio, and video, enabling agents to work with screenshots, charts, scanned documents, and voice memos.
Limitations and availability
The supplied announcement does not specify the exact number of parameters in Muse Ember.
It is available on the Meta Model API, and its performance is benchmarked against models like Claude Haiku 5.5 and GPT-6 Luna in various tasks, including the Agents' Last Exam and AutomationBench-AA. The model is intended for use in enterprise content and automation workflows, with partners reporting its effectiveness in tasks like contract review and candidate screening.
The model is priced at $0.10 per million input tokens and $0.50 per million output tokens, with no mention of additional licensing conditions.
