Meta Releases Muse Ember, a 1M-Context Model Built for High-Volume Agent Work
Meta Superintelligence Labs has released Muse Ember, its smallest and most efficient model to date. Muse Ember is a hosted multimodal reasoning model purpose-built for the repetitive steps inside agent pipelines. Meta names tool calls, coding subtasks, routing, extraction, and long-document parsing as its primary targets.
The model scores 36 on the Artificial Analysis Intelligence Index — level with Claude Haiku 5.5 at max effort. Meta puts its cost at roughly $0.14 per index task.
The release lands at a moment when agentic workloads are shifting from demos to production. As teams chain thousands of small tool calls and parsing jobs per hour, the economics of the "worker" model — rather than the frontier model — increasingly determine whether an agent product is viable.
TL;DR
- Size: 1M-token context window, up to 512K output tokens.
- Runs on: Meta's hosted infrastructure only (Meta Model API and Muse Code). Weights are private, so there is no local GPU, laptop, or quantized option.
- Performance: Ties Claude Haiku 5.5 (max effort) at 36 on the AA Intelligence Index, ahead of GPT-6 Luna (28).
- Best: AutomationBench-AA at 37.8, versus 20.2 for GPT-6 Luna and 13.9 for Claude Haiku 5.5.
- Worst: FrontierCode 2.0-AA at 27.2, behind Claude Haiku 5.5 at 34.9.
- Bottom line: Best — Haiku-level intelligence at $0.136 per task versus $0.411. Worst — closed weights and weaker performance on mergeable real-repo code.
via MarkTechPost