Meta has released Muse Ember, its smallest and most efficient AI model to date, specifically targeting routine tasks that automated agents perform daily. The system handles tool calls, minor coding jobs, request routing, data extraction, and long document processing. According to the company, Muse Ember delivers more intelligence per dollar than Claude Haiku 5.5 or GPT-6 Luna across every reasoning level. In Artificial Analysis tests, it costs approximately $0.14 per task compared to $2.39 for the larger Muse Spark 1.3, which achieves a nearly identical score of 36 versus 37. While Muse Ember leads in most internal benchmarks, the Haiku model outperforms it on real coding tasks within FrontierCode 2.0-AA, scoring 34.9 percent against Muse Ember’s 27.2 percent. The model processes text, images, audio, and video with a context window of one million tokens. It is available through the Meta Model API and Muse Code at $0.10 per million input tokens and $0.50 per million output tokens.
The launch highlights a shift toward cost-effective inference for high-volume, low-complexity operations without claiming broader capabilities. The price difference suggests significant savings for applications requiring constant token processing rather than complex reasoning. Developers can access the model immediately via existing APIs.
* Costs $0.14 per task in Artificial Analysis tests
* Handles up to one million tokens in context
* Available via Meta Model API and Muse Code
