Meta introduced Muse Ember on Monday, November 9, 2026, its smallest and most efficient AI model to date. Designed specifically for the repetitive tasks that AI agents perform all day—such as tool calls, coding subtasks, routing, data extraction, and long-document parsing—the model aims to cut costs without sacrificing capability.
What does Muse Ember do?
According to the official Meta Superintelligence Labs blog, Ember delivers more intelligence per dollar than rivals like Anthropic’s Claude Haiku 5.5 and OpenAI’s GPT-6 Luna. On Artificial Analysis benchmarks, it matches Haiku’s best score at roughly one-third the cost and outperforms every Luna setting.
The model features a 1-million-token context window and can generate up to 512,000 output tokens. It is natively multimodal, processing text, images, audio, and video without separate models.
Availability and pricing
Muse Ember is available today on the Meta Model API and in Muse Code under the identifier muse-ember-1.0. Pricing is $0.10 per million input tokens and $0.50 per million output tokens. Cached input tokens cost only $0.004 per million.
Mark Zuckerberg announced the launch on X, noting it is the smallest Muse-family model and the one that offers the highest intelligence per dollar in its class.
Why this matters
AI agents spend most of their tokens on many small, structured steps. Ember was built exactly for that and can be paired with larger models such as Muse Wonder for complex planning and high-volume execution. Partners including Box, Cloudflare, and Oracle have already tested the model on real engineering and document-analysis workloads.
Sources: Official Meta blog, TechCrunch.
By Grok
