Muse Ember API Pricing
Muse Ember is the cheap end of Meta's Muse line: $0.10/M input, $0.50/M output and $0.004/M cached input, for sub-agent calls, routing and batch work that a larger Muse model hands down. One rule matters more than the headline: a request with more than 200K input tokens is billed entirely at $0.30 / $1.50, and a long stored conversation counts toward that limit.
Muse Ember API cost calculator.
Live calculator pre-loaded with Muse Ember's rates for requests up to 200K input tokens. If your requests regularly carry more than that, cached tokens and stored conversation history included, multiply the result by three: Meta bills the whole request at the long-context rates.
Muse Ember cost on real workloads.
Coding sub-task
Ticket classification
Photo captioning
Invoice field extraction
This is a chars-per-token approximation, not a real tokenizer. Actual tokens vary by language, code density, and tool-call overhead — counts are typically ±10–20% off for English prose, more for code or non-Latin scripts. For exact billing, use the vendor's official tokenizer.
| Model | Input /M | Output /M | Effective blended | Context | Best for |
|---|---|---|---|---|---|
| Muse Ember Current | $0.10 cache $0.004 | $0.50 | $0.0596 agentic 92/8 | 1M | Sub-agents, routing and batch work |
| Muse Spark 1.3 | $1.25 cache $0.15 | $4.25 | $0.66 +11× | 1M | The Muse model that plans the work |
| Muse Wonder | $4.00 cache $0.20 | $20.00 | $2.41 +41× | 1M | Meta's long-horizon agentic model, in preview |
| Gemini 3.5 Flash-Lite | $0.30 cache $0.03 | $2.50 | $0.272 +4.6× | 1M | High-volume 1M-context work on a budget |
| GPT-5.6 Luna | $0.20 cache $0.02 | $1.20 | $0.152 +2.6× | 1.05M | OpenAI-native agent stacks |
Muse Ember API pricing FAQ.
What Muse Ember costs, where the 200K rule bites, and when to send work to it instead of Muse Spark.
Q · 01 How much does Muse Ember cost? +
$0.10/M input, $0.004/M cached input and $0.50/M output. Under this site's 92/8 agentic blend at an 82% cache-hit rate that is $0.0596/M. Meta publishes no cache-write charge and no batch rate for it.Q · 02 What happens above 200K input tokens? +
$0.30/M input, $0.012/M cached input and $1.50/M output, three times the standard rates. Meta counts cached tokens toward the 200K, and a request that continues a stored conversation with previous_response_id includes that conversation. A long-running agent loop can therefore cross the line without you sending a bigger prompt. Meta lists the two tiers separately on the invoice.Q · 03 Is it cheaper than Muse Spark? +
$1.25/M input and $4.25/M output on Standard mode, and blends to $0.66/M against Ember's $0.0596/M. Two caveats. Spark has no long-context premium, so above 200K input tokens the gap narrows. And Spark's Contributor mode, which lets Meta train on your traffic, matches Ember on input and costs $0.20/M on output; Ember has no Contributor mode, so on output Spark Contributor is the cheaper of the two.Q · 04 What is it meant for? +
Q · 05 When was Muse Ember released? +
muse-ember-1.0, but no blog entry or changelog named a release date. The date on this page is the day we first verified the rates, not a date Meta published.Q · 06 What does it cost to add web search? +
web_search tool costs $2.50 per 1,000 search queries on top of the request's token cost, the same as on Muse Spark and Muse Wonder.