LLM API pricing / Meta
What the Meta Muse API costs, model by model
Short answer. Meta lists 4 models with a per-token price on its API. Input: from $1.25 (Muse Spark 1.1, Muse Spark 1.2 and Muse Spark 1.3) to $4.00 (Muse Wonder) per 1M tokens. Output: from $4.25 (Muse Spark 1.1, Muse Spark 1.2 and Muse Spark 1.3) to $20.00 (Muse Wonder). Read on Meta's own pages between October 26 and October 27, 2026.
In plain words. You pay for the tokens you send (input) and the tokens the model writes (output). One request of 10,000 tokens in and 1,000 out costs $0.0168 on Muse Spark 1.1, Muse Spark 1.2 and Muse Spark 1.3 and $0.06 on Muse Wonder.
Current models
4 models, cheapest input first, verified between October 26, 2026 and October 27, 2026. The last column is the cost of one request of 10,000 input and 1,000 output tokens at these prices, with no cache.
| Model | Input $/1M | Output $/1M | Cached input | Context | Example request | Verified |
|---|---|---|---|---|---|---|
| Muse Spark 1.1 | $1.25 | $4.25 | $0.15 | 1.0M | $0.0168 | 2026-10-27 |
| Muse Spark 1.2 | $1.25 | $4.25 | $0.15 | 1.0M | $0.0168 | 2026-10-26 |
| Muse Spark 1.3 | $1.25 | $4.25 | $0.15 | 1.0M | $0.0168 | 2026-10-26 |
| Muse Wonder preview | $4.00 | $20.00 | $0.20 | 1.0M | $0.06 | 2026-10-27 |
How Meta describes its models
- Muse Spark 1.2:
Muse Spark 1.2 is optimized for real coding workflows, with higher first-attempt accuracy and more reliable tool calling.
(dev.meta.ai, read September 30, 2026) - Muse Spark 1.3:
Muse Spark 1.3 is trained for agentic workflows and optimized for competitive coding performance.
(dev.meta.ai, read September 30, 2026) - Muse Wonder:
Frontier coding, long-running agents, and advanced computer use. Plans long tasks, delegates to sub-agents, and recovers from errors.
(dev.meta.ai, read October 14, 2026)
How Meta prices its models
- 4 of the 4 models publish a cached-input price, at 5% or 12% of the input price.
What Meta's own pages say
Preview
- Muse Wonder:
Muse Wonder is in preview.
(dev.meta.ai, read October 26, 2026)
Contributor mode
- Meta's pricing and rate limits page:
Contributor mode offers discounted pricing or more usage in exchange for permission to use your content — the prompts, code, files, and context you send, plus the model responses you get back — to train future Meta models.
(dev.meta.ai, read October 26, 2026) - Meta's pricing and rate limits page:
In Standard mode, you pay standard prices and your content is not used to train Meta models.
(dev.meta.ai, read October 26, 2026) - Meta's pricing page, Muse Spark table (Standard mode | Contributor mode, per 1M tokens):
Cached input | $0.15 | $0.002 |
(dev.meta.ai, read October 26, 2026) - Meta's pricing page, Muse Spark table (Standard mode | Contributor mode, per 1M tokens):
Input | $1.25 | $0.10 |
(dev.meta.ai, read October 26, 2026) - Meta's pricing page, Muse Spark table (Standard mode | Contributor mode, per 1M tokens):
Output | $4.25 | $0.20 |
(dev.meta.ai, read October 26, 2026) - Muse Wonder:
Contributor mode isn't available for Muse Wonder on Model API.
(dev.meta.ai, read October 26, 2026) - Muse Spark 1.1:
Muse Spark 1.1 is Standard mode only.
(dev.meta.ai, read October 26, 2026)
Long prompts
- Meta's pricing and rate limits page:
you pay the same rate whether your context window is mostly empty or almost full.
(dev.meta.ai, read October 26, 2026)
Web search
- Meta's pricing and rate limits page:
$2.50 per 1,000 search queries, in addition to the request's token cost.
(dev.meta.ai, read October 26, 2026)
Retirements
Retired: Llama API (Meta) (July 6, 2026).
- July 6, 2026: Closure of the Llama API. Meta no longer serves Llama models through its own API; replaced by the Meta Model API with the Muse family source
Recorded price changes
- October 14, 2026 · correction of our own reading. Correction of our own reading, not a price change by Meta. Since September 16, 2026 AI Signal showed Muse Spark 1.1 as 'no published price', because its own model page carries a benchmark table and no pricing table. On October 14, 2026 Meta's 'Pricing and rate limits' page, which AI Signal had not read, lists muse-spark-1.1 in Standard mode next to 1.3 and 1.2 ('Muse Spark 1.1 is Standard mode only.') at $1.25 input, $0.15 cached input and $4.25 output per million tokens. The row now carries that price. When Meta began listing 1.1 on that page is not known, so the date here is the day it was read. source
- September 16, 2026 · correction of our own reading. Correction of our own reading, not a price change by Meta. From September 11 to September 16, 2026 AI Signal showed Muse Spark 1.3 and Muse Spark 1.2 as 'no published price', because Meta's model pages serve their 'Models and pricing' table only to a browser, and the HTTP client AI Signal verifies with received a JavaScript shell with no pricing text. Read by hand in a browser on September 16, 2026, each page lists two SKUs per model: 'Not used to improve our products' at $1.25 input, $0.15 cached input and $4.25 output per million tokens, and a 'contributor' SKU, 'Used to improve our products', at $0.10 / $0.002 / $0.20, conditional on letting Meta use the traffic to improve its products. The rows now carry the first as the list price, with the contributor price quoted in their notes; Meta changed nothing. Muse Spark 1.1's own model page, read the same day in a browser, carries a benchmark table and no pricing table, and neither of the other two tables lists it, so that row still shows no published price. source
Analysis that cites Meta models
- AI Pricing This Week: Gemini 4 Argon, Muse Wonder, Oct 12-18 · names 2 Meta models
- Gemini 4 Argon vs Muse Wonder: Same Price, Until Nov 30 · names 2 Meta models
- The Second Price List: What Batch, Flex, Priority and Data-Sharing Discounts Cost Across Seven Providers · names 2 Meta models
- Prompt Caching Is Not 10% Everywhere: What a Cache Hit Costs Across Eight Providers · names 2 Meta models
- Same List Price, Different Bill: Five Pairs of LLM APIs Where the Workload Decides What You Pay · names 1 Meta model
FAQ
What is the cheapest Meta Muse model on the API?
Muse Spark 1.1, Muse Spark 1.2 and Muse Spark 1.3, each at $1.25 per 1M input tokens and $4.25 per 1M output tokens, as Meta's pricing page printed it on October 27, 2026.
Is Muse Wonder generally available?
Muse Wonder: “Muse Wonder is in preview.” Read on October 26, 2026.
What is Contributor mode on Meta's Model API?
Meta's pricing and rate limits page: “Contributor mode offers discounted pricing or more usage in exchange for permission to use your content — the prompts, code, files, and context you send, plus the model responses you get back — to train future Meta models.” “In Standard mode, you pay standard prices and your content is not used to train Meta models.” Meta's pricing page, Muse Spark table (Standard mode | Contributor mode, per 1M tokens): “Cached input | $0.15 | $0.002 |”. “Input | $1.25 | $0.10 |”. “Output | $4.25 | $0.20 |”. Muse Wonder: “Contributor mode isn't available for Muse Wonder on Model API.” Muse Spark 1.1: “Muse Spark 1.1 is Standard mode only.” Read on October 26, 2026.
Do Muse prices change for long prompts?
Meta's pricing and rate limits page: “you pay the same rate whether your context window is mostly empty or almost full.” Read on October 26, 2026.
Does web search cost extra on Meta's Model API?
Meta's pricing and rate limits page: “$2.50 per 1,000 search queries, in addition to the request's token cost.” Read on October 26, 2026.
How we checked
- Each price was read on Meta's pricing page (dev.meta.ai) or on a model page it links. The Verified column links to that page and gives the day.
- Quotes are copied as Meta's pages print them, with the day they were read. A script checks every week that the model descriptions and the quotes under “What Meta's own pages say” are still on their pages.
- The example request is worked out from the per-1M prices: 10,000 input tokens and 1,000 output tokens, no cache, no batch discount, no long-prompt tier.
Full method: methodology. All providers: the full pricing table.