Use Meta AI assistant to get things done, create AI-generated images for free, and get answers to any of your questions. Meta AI is built on Meta's latest Llama large language model and uses Emu.
This is the 7th launch from Meta AI. View more
Muse Ember by Meta
Launched this week
Muse Ember is Meta's smallest multimodal reasoning model, built to handle the repetitive work behind AI agents. From tool calls and coding subtasks to document extraction and routing, Ember delivers efficient execution at scale. It supports text, images, audio, and video, with a 1M-token context window and pricing starting at $0.10 per million input tokens. Pair it with larger models for planning, and let Ember handle the volume. Available through Meta Model API and Muse Code.
Payment Required
Launch tags:Artificial Intelligence
Launch Team
Render Sandboxes: a computer for every agentRun untrusted code safely. Idle sandboxes cost nothing
Promoted
Meta just released Muse Ember, its smallest and cheapest Muse model yet.
What's interesting is that it's designed specifically for the work AI agents repeat all day: tool calls, coding subtasks, routing, document processing, and so on.
The idea is to let a larger model handle the difficult planning, then use Ember to execute the smaller tasks at scale.
It supports text, images, audio, and video, has a 1M-token context window, and starts at just $0.10 per million input tokens.
Pretty interesting option for anyone building agents where inference costs add up quickly.
The split you describe — a bigger model plans, Ember executes the repetitive steps — is the part I'm most interested in. At $0.10 per million input tokens with a 1M-token window, the math for high-volume agent loops actually starts to work. Curious how it holds up on tool-call reliability over long runs, since that's where the small models usually start to drift for me.