Muse Ember

Muse Ember is Meta's smallest and most efficient model, built for the repetitive work that agents perform all day. Key capabilities: • Tool calls and function execution • Coding subtasks and code assistance • Request routing and task dispatching • Data extraction from structured and unstructured sources • Long document parsing with a 1M-token context window • Multimodal input across text, images, audio, and video • Low-cost API access for high-volume workloads. Muse Ember is designed to deliver strong reasoning per dollar. Per Meta, it matches the best score of Claude Haiku 5.5 on a leading independent benchmark leaderboard at about a third of the cost, and it outperforms every GPT-6 Luna reasoning setting in the same comparison. Few models on that leaderboard score higher for the price per task. The 1M-token context window lets developers process long contracts, research reports, support logs, and multi-file codebases in a single pass. Because the model handles text, images, audio, and video, a single endpoint can cover transcription review, visual document reading, and video summarization. Pricing is $0.10 per million input tokens and $0.50 per million output tokens, which keeps repeated calls economical when an agent makes thousands of small decisions per session. Muse Ember is available through the Meta Model API and Muse Code. Developers building agents, routing layers, and extraction pipelines will find it well suited to subtasks that need quick, reliable responses rather than maximum model size. Engineering teams can use it for code helpers, document processing backends, and cost-sensitive production deployments where speed and efficiency matter most.