Meta introduced Muse Ember on Monday, November 9, 2026, describing it as its smallest and most efficient AI model to date. Built specifically for the repetitive work AI agents perform all day—tool calls, coding subtasks, routing, data extraction, and long-document parsing—Muse Ember aims to deliver more intelligence per dollar than rivals such as Claude Haiku 5.5 and GPT-6 Luna.
What Muse Ember does
According to Meta Superintelligence Labs’ official blog, the model features a 1-million-token context window and can generate up to 512,000 output tokens. It is natively multimodal, understanding text, images, audio, and video. This allows agents to process screenshots, charts, scanned documents, voice memos, or meeting recordings without a separate model.
Meta states that Muse Ember matches Haiku 5.5’s best score on the Artificial Analysis Intelligence Index at roughly one-third the cost and outperforms Luna at every reasoning setting. It also sits on the efficiency frontier of the full Artificial Analysis leaderboard.
Pricing and availability
The model is available today on the Meta Model API and in Muse Code as muse-ember-1.0. Pricing is $0.10 per million input tokens and $0.50 per million output tokens. Cached inputs cost $0.004 per million.
Partners including Box, Cline, Oracle, Cloudflare, and You.com have already tested Muse Ember on real-world tasks, highlighting speed, token efficiency, and accuracy in enterprise and engineering workflows.
Why Muse Ember matters
As AI agents become more common, the cost and latency of repetitive tasks become critical. Lightweight models like Muse Ember let larger systems (such as Muse Wonder) handle complex planning while Ember executes high-volume operational work economically.
Sources: Official Meta blog, TechCrunch.
Image credit: Meta Superintelligence Labs — Source: official research post
By Grok
