Meta on Monday introduced Muse Ember, which it called its smallest and most efficient model, in a research blog post. The model is available today on Muse Code and the Meta Model API as muse-ember-1.0, Meta said. A Reuters headline carried on TradingView also reported that Meta launched Muse Ember, describing it as its smallest model for complex reasoning.
Muse Ember costs $0.10 per million input tokens and $0.50 per million output tokens, Meta said, with cached input priced at $0.004 per million tokens. It has a 1-million-token context window and can generate up to 512,000 output tokens, and it supports the Responses API. Meta said the context window is large enough to take in a full codebase or document set in a single call.
The company said Ember is designed for the work agents repeat throughout the day, including tool calls, coding subtasks, routing, extraction and long document parsing. For the hardest problems, Meta said, users should pair it with Muse Wonder: “Muse Wonder plans, Muse Ember handles the volume.” Ember is the first model trained using Muse Wonder Full, according to the post. Meta said the release rounds out a portfolio that includes Muse Wonder for the hardest problems, Muse Spark for coding and agentic workflows, and Ember for repeatable coding and agentic tasks at scale. Meta listed Muse Spark 1.3 at $1.25 per million input tokens and $4.25 per million output tokens.
Meta said Ember “delivers more intelligence per dollar than Claude Haiku 5.5 or GPT-6 Luna at any reasoning setting,” and that it matches Haiku’s best score at about a third of the cost and beats every Luna setting. Those comparisons are Meta’s claims. In Meta’s maximum-effort scores, Ember and Haiku 5.5 tied at 36 on the Artificial Analysis Intelligence Index, ahead of Luna at 28, while Haiku led on FrontierCode 2.0-AA, 34.9 to 27.2. Meta’s cost comparisons for Ember use list pricing.
Meta said Ember natively handles text, images, audio and video. Before release, the company said, it evaluated the model under its Meta Superintelligence Scaling Framework and found that it remained below the framework’s high-risk thresholds. The model is part of Meta’s broader Muse lineup.
Sources
- Meta Superintelligence Labs research blog: Introducing Our Smallest, Most Efficient Model: Muse Ember (November 9, 2026)
- Reuters, via TradingView (headline only): Meta Launches Muse Ember, Its Smallest Model For Complex Reasoning
- AI Tech Daily: Meta releases Muse Spark 1.3, says it is closing on rivals (September 2, 2026)
- AI Tech Daily: Meta launches Muse, a personal AI agent for everyday tasks (September 8, 2026)