Meta Superintelligence Lab released Muse Ember on Monday, a lightweight AI model designed for high-volume agentic tasks such as coding subtasks, tool calls, routing, and long document parsing. The company stated the model is optimized for complex reasoning execution and repetitive workflows.
At a price point of 10 cents per million input tokens, Ember is positioned as a cost-efficient alternative to frontier models. Meta cited Artificial Analysis’s composite benchmark score to claim the model is cheaper and smarter than Anthropic’s Haiku and OpenAI’s Luna. The company also noted that Ember is cost-competitive with open source models including DeepSeek V4.1 Flash and GLM 5.3 Flash.
Meta intends for Ember to function within an agentic harness alongside larger models like Muse Wonder. In this configuration, a heavier model establishes rules for specific issues while the lighter model applies those rules across large volumes of cases. Meta described the relationship by stating, “Muse Wonder plans, Muse Ember handles the volume.”
The model is trained to process multiple media types, including text, images, and audio, allowing it to handle inputs such as screenshots, charts, scanned documents, voice memos, meeting recordings, and video without requiring a separate model.
Ember is the fourth release in the Muse brand, which includes the Spark mid-weight model, the open source Glimmer, and the Wonder frontier model. This follows the September release of the Muse Agent consumer product.
