Meta launches Muse Ember, a small AI model for agent tasks

Muse Ember slots into Meta's model lineup as the smallest and most efficient option, aimed squarely at the repetitive work that AI agents perform continuously. The pricing reflects that focus: at $0.10 per million input tokens and $0.50 per million output tokens, it is built for high-volume, low-complexity calls rather than heavyweight reasoning.

Meta's pitch centers on cost efficiency. It says Muse Ember delivers more intelligence per dollar at every reasoning level than Claude Haiku 5.5 or GPT-6 Luna. In the Artificial Analysis test, Muse Ember costs about $0.14 per task while the larger Muse Spark 1.3 costs $2.39 for a nearly identical score of 36 versus 37, underscoring how much of the model's value proposition rests on economics rather than raw capability.

The trade-off appears on coding. Haiku outperforms Muse Ember on real coding tasks in FrontierCode 2.0-AA, scoring 34.9 percent to Muse Ember's 27.2 percent. That gap suggests Meta is not positioning Muse Ember as a general-purpose coding model but as a cheap engine for routing, extraction, and tool calls — the unglamorous but constant work that keeps agent systems running.