Meta Launches Muse Ember at 10 Cents: Cheapest AI Model for Agents
Meta has launched Muse Ember, its smallest AI model yet, at $0.10 per million input tokens — built for the repetitive work AI agents do all day.
Image: Wikimedia Commons / Nokia621
MENLO PARK, California, Nov. 9 — Meta launched Muse Ember on Monday, its smallest and most efficient artificial intelligence model, priced at $0.10 per million input tokens and $0.50 per million output tokens for the repetitive work AI agents perform all day.
The release arrives the same week as Anthropic's Claude Fable 5.2 and Google's new TPU systems, and it draws a clear line: flagships keep growing more capable, while a cheaper tier of executor models runs agents at scale.
Key Takeaways
- Muse Ember costs $0.10 per million input tokens and $0.50 per million output tokens, with cached input at $0.004.
- The model carries a 1-million-token context window and can generate up to 512,000 output tokens.
- Artificial Analysis scores Ember at 36 on its Intelligence Index, matching Claude Haiku 5.5, at about $0.14 per task against $2.39 for Muse Spark 1.3.
- Ember is available now as muse-ember-1.0 on Muse Code and the Meta Model API; Meta has not released the weights.
How Much Does Meta Muse Ember Cost? Full Pricing and Token Limits
Ember's headline price is $0.10 per million input tokens, $0.004 cached input, and $0.50 per million output tokens, for requests up to 200,000 input tokens. Above that, the whole request bills at triple the base rate — $0.30, $0.012, and $1.50 — so one large document can triple a job's cost.
| Request size | Input (per 1M tokens) | Cached input (per 1M tokens) | Output (per 1M tokens) |
|---|---|---|---|
| Up to 200,000 input tokens | $0.10 | $0.004 | $0.50 |
| Over 200,000 input tokens | $0.30 | $0.012 | $1.50 |
At the standard rate, Ember matches Claude Haiku 5.5 and OpenAI's GPT-6 Luna, both also $0.10 and $0.50. Meta's larger models cost more: Muse Spark 1.3 is $1.25 and $4.25 per million tokens, and flagship Muse Wonder is listed at $4 and $20.
Muse Ember vs Claude Haiku 5.5 and GPT-6 Luna: Price and Performance
Artificial Analysis, the independent benchmarking firm, puts Ember at 36 on its Intelligence Index, level with Claude Haiku 5.5. One index task costs about $0.14 for Ember, against $2.39 for Muse Spark 1.3, which scores 37. Ember does not outthink the flagships — it matches its small-model rivals for a fraction of the cost of routine work.
The contrast with Anthropic's new flagship shows the split. Claude Fable 5.2, launched the same day, scores 65 on the same index and costs $8 per million input tokens and $40 per million output tokens — 80 times Ember's input price. Fable 5.2 is built for the hardest coding and science work; Ember is built for the volume underneath it.
What Is Muse Ember Built For? Agent Tasks, Limits and Availability
Meta positions Ember for the jobs agents repeat: tool calls, coding subtasks, request routing, data extraction, and parsing long documents. It understands text, images, audio, and video, and it is the first model Meta says was trained using Muse Wonder Full. Before release, Meta tested it for chemical and biological risks, cybersecurity, and loss-of-control scenarios.
There are limits buyers should know. Meta has not released Ember's weights or its parameter count, so teams cannot run it on their own hardware. The model is available only through Muse Code and the Meta Model API, as muse-ember-1.0. And the low headline price only holds up to 200,000 input tokens; anything larger bills at triple the rate.
Frequently Asked Questions
How much does Meta Muse Ember cost per million tokens?
Ember costs $0.10 per million input tokens, $0.004 cached, and $0.50 per million output tokens for requests up to 200,000 input tokens. Above that, the whole request bills at $0.30, $0.012, and $1.50.
Is Muse Ember better than Claude Haiku 5.5?
On Artificial Analysis benchmarks, Ember scores 36 on the Intelligence Index, matching Claude Haiku 5.5, at the same $0.10 and $0.50 list prices. Ember is built for routine agent work; larger models score higher on the hardest reasoning tasks.
Where can developers use Muse Ember?
Ember is available now as muse-ember-1.0 on Muse Code and the Meta Model API. Meta has not released the weights, so it cannot be downloaded or run on developers' own hardware.
What is Muse Ember used for?
Meta built Ember for the repetitive work AI agents do all day: tool calls, coding subtasks, request routing, data extraction, and parsing long documents. It carries a 1-million-token context window and generates up to 512,000 output tokens.
Meta has not said whether it will release Ember's weights or extend it beyond Muse Code and the Meta Model API.
