Meta launches Muse Ember for low-cost agent work
Meta launched Muse Ember, a model designed for recurring agent tasks such as tool calls, routing, data extraction, coding subtasks, and long-document parsing. It supports text, images, audio, and video, with a 1 million-token context window and up to 512,000 output tokens, and is available through the Meta Model API and Muse Code at $0.10 per million input tokens and $0.50 per million output tokens. Meta says it matches Claude Haiku 5.5’s best Artificial Analysis score at roughly one-third the estimated cost per task and outperforms GPT-6 Luna, although Claude Haiku 5.5 scored higher on FrontierCode 2.0-AA. Why it matters: Muse Ember gives developers a low-priced option for repeated, token-heavy agent operations, while its million-token context and multimodal inputs broaden the kinds of workloads it can handle. What do you think?
Open the headline: source, AI briefing and more →Comments
…