Meta Muse Ember is now available in FastMetal chat
Muse Ember is live in FastMetal chat. Meta released it on November 9, 2026, as its smallest and most efficient model, meant for the high-volume agent work that bigger models are overkill for. It sits below Muse Wonder and Muse Spark 1.3, which we also serve.
What Meta built it for
Meta's developer page lists coding subtasks, tool calls, routing, document processing and extraction. Meta suggests pairing it with Muse Wonder or Muse Spark: the bigger model plans, Ember does the repeated steps.
On Meta's own comparison table, Ember scores 40.9 on Terminal-Bench 4.0 against 39.2 for Claude Haiku 5.5, but trails it on FrontierCode 2.0, 27.2 against 34.9. These are Meta's figures. If you are choosing between small models, Claude Haiku 5.5 and GPT-6 Luna are the obvious ones to compare it with.
What we checked before turning it on
- Image input works: it named the colors in a test image correctly.
- PDF input works: a PDF attached as-is was read back correctly.
- It always reasons before answering, and those reasoning tokens bill as output tokens.
- The context window is 1,048,576 tokens.
Pricing
Muse Ember bills per million input and output tokens, with a separate rate for cached input. Current rates are on the Muse Ember model page, and billing is explained on /pricing. Usage is prepaid, so you can't be charged past your balance.
How to use it
- Create a FastMetal account.
- Add credit.
- Open chat, pick Muse Ember from the model list, and ask. You can attach images and PDFs to a message.
Compare real answers first
The model page shows Muse Ember's actual answers to FastMetal's 43 shared prompts (summarization, bug fixes, translation and more), next to Muse Wonder and every other model we serve. The full catalog is at /models.