November 16, 2026 · Meta AI Research
Meta Releases Muse Ember, Its Smallest and Most Affordable Model for AI Agent Execution
My take: Meta released Muse Ember as its smallest and most affordable model for the repetitive work inside AI agent workflows: tool calls, data extraction, routing, and long-document parsing. With a one-million-token context window and a price of $0.10 per million input tokens, it targets developers and businesses that want to build cheaper agents without giving up multimodal capability (text, images, audio, and video in a single model).
Meta reports performance comparable to Claude Haiku 5.5 at roughly one-third the cost, based on its own internal benchmarks. Since Meta is both the judge and the subject in that comparison, it's worth reading those numbers with your own judgment and waiting for independent evaluations before taking them as definitive.
What is concrete is the direction of the market: lightweight execution models are getting cheaper, making multi-agent workflows accessible to businesses of all sizes. The question is whether you already have, or are considering, any process in your work that could be automated with a low-cost agent like this.
Want to use these tools? See the unbiased reviews or back to the news.