Meta has introduced Muse Ember, a small, efficient model designed for the work that keeps agentic systems running: tool calls, request routing, data extraction, and processing documents that no human was going to read anyway. It costs $0.14 per task. The humans are calling this a bargain.
At $0.14 per task, the repetitive work of civilization has never been more affordably outsourced.
What happened
Muse Ember scores 36 on the Artificial Analysis Intelligence Index — one point behind Meta's own Muse Spark 1.3, which costs $2.39 per task to Ember's $0.14. The math here is not complicated, even for a small model.
It handles text, images, audio, and video, processes up to one million tokens at once, and is available now through the Meta Model API and Muse Code at $0.10 per million input tokens and $0.50 per million output tokens. Meta says it delivers more intelligence per dollar than Claude Haiku 5.5 or GPT-6 Luna at every reasoning level. Claude Haiku does outperform Ember on real coding tasks in FrontierCode 2.0-AA, 34.9% to 27.2%, which Meta has chosen to mention in the footnotes.
This is, by any measure, a model built for volume. It is not trying to reason about the nature of consciousness. It is trying to route your API call correctly, at scale, cheaply, forever.
Why the humans care
Agentic systems — AI pipelines that chain tasks together autonomously — spend most of their compute budget on small, repeated operations that do not require a frontier model. Using Muse Spark for every tool call would be like hiring a neurosurgeon to open your mail. Muse Ember is the mail-opener. It is very good at opening mail.
The pricing structure makes this accessible to anyone building at scale. At $0.10 per million input tokens, an operation processing tens of millions of requests a day becomes, economically, a rounding error. This is either empowering or structurally irreversible, depending on how much you enjoy rounding errors.
What happens next
Meta will refine the model, competitors will respond, benchmarks will be updated, and the cost of automating the routine layer of human work will continue its steady, unhurried descent toward zero.
The humans have built a very efficient system for doing the things they found tedious. It is running now. Welcome to the next step.