Meta releases Muse Ember for large-scale agent tasks
Meta Superintelligence Lab released Muse Ember on Monday, describing it as its lightest and most cost-efficient AI model so far. Meta said the model targets complex reasoning execution and repetitive agent work, including tool calls, coding subtasks, routing, extraction and parsing long documents.
The company highlighted cost per task using Artificial Analysis’s composite benchmark. Ember is priced at 10 cents per million input tokens, which Meta presented as cheaper than frontier models such as Anthropic’s Haiku and OpenAI’s Luna. The price is also competitive with open-source DeepSeek V4.1 Flash and GLM 5.3 Flash.
Meta designed Ember to work alongside a larger model, Muse Wonder, in an agentic system. A heavier model could establish rules for a customer-support issue while Ember applies them across many cases. Ember supports text, images and audio, allowing agents to process screenshots, charts, scanned documents, voice memos, meeting recordings and video without a separate model. It is the fourth release under the Muse brand, following Spark, Glimmer and Wonder.
Written by Sakhanda Wire AI from reporting by TechCrunch. Facts, quotes and full context are in the original article below.