Meta Unveils Muse Ember: Lightweight, Cost‑Effective AI Model for Agentic Workloads
A New Lightweight Model for Complex Agent Tasks
Meta Superintelligence Lab launched Muse Ember on Monday, describing it as the lab’s lightest and most cost‑efficient AI model to date. According to the release, Ember is built for the intricate reasoning required in everyday agent activities such as tool calls, coding subtasks, routing, extraction, and parsing long documents.
Cost Efficiency and Benchmark Performance
Meta highlighted Ember’s cost‑per‑task by referencing Artificial Analysis’s composite benchmark score, positioning Ember as both cheaper and more capable than frontier rivals like Anthropic’s Haiku and OpenAI’s Luna.
Meta emphasized that Ember’s pricing of 10 cents per million input tokens makes it significantly less expensive than Haiku and Luna, while remaining cost‑competitive with open‑source options such as DeepSeek V4.1 Flash and GLM 5.3 Flash.
Integration with Muse Wonder for Scalable Agentic Workflows
The model is intended to operate alongside a larger counterpart, such as Meta’s Muse Wonder, within an agentic harness. In a customer‑support scenario, a heavier model could define handling rules, and Ember would apply those rules across tens of thousands of cases.
Meta explained, “Muse Wonder plans, Muse Ember handles the volume.”
Multimodal Capabilities Enable Diverse Input Types
Designed for volume‑heavy applications, Ember was trained to process multiple media types, including text, images, and audio. This multimodal training allows agents to interpret screenshots, charts, scanned documents, voice memos, meeting recordings, and video without needing separate models.
Meta stated, “Agents can read screenshots, charts, and scanned documents, or work from voice memos, meeting recordings, and video without a separate model.”
Meta’s Expanding Muse AI Portfolio
Muse Ember marks the fourth release under Meta’s Muse brand, following the mid‑weight Spark (April), its open‑source twin Glimmer (August), and the frontier Wonder model (October). The consumer‑focused Muse Agent debuted in September to unexpected success, with new features like Spaces and video chat introduced earlier this month.








