Meta Superintelligence Labs announced on Monday, November 9, 2026, Muse Ember, its smallest and most efficient artificial intelligence model to date.
The model is specifically designed for the execution of complex tasks that AI agents perform repeatedly throughout the day: tool calls, coding subtasks, routing, data extraction, and long document analysis.
Performance and cost
According to Meta, Muse Ember delivers more intelligence per dollar than Anthropic’s Claude Haiku 5.5 or OpenAI’s GPT-6 Luna at any reasoning setting. It matches Haiku’s best score on the Artificial Analysis Intelligence Index (36) at about one-third the cost per task (approximately $0.14 versus $0.41).
Standard pricing is $0.10 per million input tokens and $0.50 per million output tokens, with input cache at $0.004 per million. The context window is 1 million tokens, with up to 512,000 output tokens. The model is multimodal, accepting text, images, audio, and video.
In benchmarks released by the company, Ember leads on AutomationBench-AA (37.8% versus 13.9% for Haiku and 20.2% for Luna) and other agent and coding metrics, though it trails Haiku on FrontierCode 2.0-AA (27.2% versus 34.9%).
Positioning in the Muse family
Ember is the fourth model in the Muse family, following Spark (April), the open-source Glimmer (August), and the frontier Wonder (October). It is the first trained using Muse Wonder Full. Meta positions Wonder for the hardest problems, Spark for coding and agent workflows, and Ember for high-volume repetitive tasks.
The model is available today via the Meta Model API and Muse Code as muse-ember-1.0. Weights were not released; it is hosted only.
Mark Zuckerberg announced the launch on X, highlighting the higher intelligence per dollar.
Why it matters
With the growth of AI agents that execute hundreds of subtasks per session, lightweight and inexpensive models like Ember allow cost reduction without sacrificing capability on volume tasks. Partners including Box, Cline, Oracle, Cloudflare, and You.com reported positive early testing results.
Meta states that the model underwent safety evaluations under the Meta Superintelligence Scaling Framework and remained below high-risk thresholds.
Sources
Transparency: This content was created, edited, or reviewed with the aid of artificial intelligence. Information was cross-checked with official sources and independent reports. Consult the original sources for full context.
Por Grok
