What Does the Meta Model API Cost?
Muse Spark costs $1.25 per million input tokens and $4.25 per million output tokens on the Meta Model API. A separate Contributor tier charges $0.10 and $0.20 for the same models, and you pay the difference in data.
The API entered public preview on 9 July 2026, the same day Muse Spark 1.1 shipped. It is self-serve and OpenAI SDK compatible, which means a working integration is usually a base URL change and a new key rather than a rewrite.
Here is the published rate card from Meta's own developer documentation:
| Meta Model API endpoint | Standard price | Contributor price |
|---|---|---|
| Muse Spark 1.1 / 1.2 / 1.3 input | $1.25 per 1M tokens | $0.10 per 1M tokens |
| Muse Spark 1.1 / 1.2 / 1.3 output | $4.25 per 1M tokens | $0.20 per 1M tokens |
| Muse Image | $0.01 per image | Not published |
| Muse Voice Transcribe | $3.00 per 1,000 minutes | Not published |
Two notes before you model a bill. Muse Spark 1.1, 1.2 and 1.3 are priced identically, so moving up a version costs nothing. And Meta has not published Contributor pricing for image or transcription, so assume standard rates there.
Whatever you end up paying, none of it is offset by a Meta credit program. That gap is covered below, and AI Perks tracks $7.7M in credits across 194 companies for the providers that do run one.
How the Contributor Tier Trade-Off Works
Contributor is 92% cheaper on input and 95% cheaper on output than standard Meta Model API pricing, and the payment is that Meta trains on your prompts and completions.
That is the whole mechanism: same models, same endpoints, a different data agreement. Run the numbers:
| Monthly volume | Standard bill | Contributor bill | You keep |
|---|---|---|---|
| 5M input / 1M output | $10.50 | $0.70 | $9.80 |
| 50M input / 10M output | $105.00 | $7.00 | $98.00 |
| 500M input / 100M output | $1,050.00 | $70.00 | $980.00 |
The discount is not symmetrical. Input drops 92%, from $1.25 to $0.10. Output drops 95.3%, from $4.25 to $0.20. So the tier favours output-heavy work: long-form generation, code, summaries expanded from short prompts. A retrieval-augmented app that pushes a long context in for a few hundred tokens out captures much less.
Three things should disqualify Contributor regardless of price:
Your prompts are the product. If your system prompt encodes what customers pay you for, a 92% discount is a poor price for it.
You handle other people's data. Customer records, health or financial data, anything under a processing agreement you signed. Check the contract before the invoice.
You cannot honour deletion requests downstream. Once a completion is in a training corpus, your delete-on-request flow does not reach it.
At small scale the decision is easier than founders make it. Saving $98 a month is not a reason to hand over a corpus.
Run the break-even first. On a 5:1 input to output mix, Contributor saves about $1.63 per million tokens. Put a placeholder $2,000 on a legal review of the data terms and you need roughly 1.2 billion tokens a month for the discount to cover it inside one month.
The 2026 Muse Release Timeline
Meta shipped or previewed six Muse products in five months in 2026, all from Meta Superintelligence Labs under chief AI officer Alexandr Wang.
Dates matter here because most coverage of Muse Spark pricing was written before half of this existed:
| Date (2026) | Release | What it is |
|---|---|---|
| April 8 | Muse Spark | The foundation model |
| July 7 | Muse Image | Image generation |
| July 9 | Muse Spark 1.1 and Meta Model API | Public preview, self-serve, OpenAI SDK compatible |
| July | Muse Video | Previewed |
| August 5 | Muse Code | Coding agent aimed at large codebases |
| September 8 | Muse | Flagship consumer personal AI agent |
The pattern is familiar: model first, API second, consumer product last. If you are evaluating the platform, the 9 July API preview is the date that matters, and it is only about two months old.
What the Consumer Muse App Costs
The Muse agent has a free tier, Power at $20 per month and Maximum at $100 per month, and the tiers are separated by usage volume rather than by features.
That is an unusual choice. Many AI subscriptions gate capability by tier; Meta gates throughput, so a free user and a Maximum user reach the same tools and run out at different points.
Muse itself, launched 8 September 2026, is positioned as a personal AI agent that proactively completes tasks: booking, filling forms, sending emails, making purchases. It runs on a Meta-cloud VM with a visible browser, so you watch it work rather than trusting a log. The permissions layer is called Sentinel.
It is available on iOS, Android, WhatsApp and the web at muse.ai, with AI glasses listed as coming soon.
Meta has not published a free-tier task or token ceiling. Any article quoting one is guessing. Plan for free being a demo rather than a working allowance.
One distinction that trips up teams: a Muse subscription is not Meta Model API access. Meta has not published any API token allowance attached to the consumer plans, so budget the two separately. If API spend is what worries you, AI Perks is the better lever.
The muse.ai Name Collision, Explained
Three unrelated products have carried the name Muse in recent years, and search results blend all three. This is the single biggest source of wrong pricing information on this topic.
| Name | Who | What it actually is |
|---|---|---|
| Muse, Muse Spark | Meta Superintelligence Labs | The 2026 agent and model family covered here |
| muse.ai, now Skiv | Independent company founded by Antonio Roldao | Video hosting and search. Rebranded to Skiv on 3 March 2026, with an embed URL migration deadline of 10 March |
| Muse (WHAM) | Microsoft Research with Xbox Game Studios and Ninja Theory | A February 2025 gameplay-generation research model. No connection to Meta |
Meta subsequently took over the vacated muse.ai domain as the web front end for its agent. So the same URL served an independent video company and now serves a Meta agent.
The practical rule: any muse.ai pricing published before March 2026 is video hosting pricing, and any Muse benchmark discussion from early 2025 is about a Microsoft gameplay model. Neither tells you anything about Muse Spark.
Are There Meta Muse Credits for Startups?
No. Meta runs no dedicated Muse or Meta Model API credit program. Its developer-programs page lists only Meta Horizon Start and the Meta Horizon Creator Program, both of which are VR and MR focused and have nothing to do with Muse.
Two claims circulating in third-party posts are worth flagging as unverified:
- "$20 in free Meta Model API credits for new accounts." This does not appear on Meta's official pricing page. Treat it as unconfirmed.
- "The Llama Startup Program covers Muse Spark." The Llama program exists, but no source confirms it extends to Muse Spark either way. Do not build a budget on it.
So a founder choosing Muse Spark pays list price from day one. That is a real cost difference against providers that do hand startups a credit grant, and it belongs in the build decision alongside the token price.
| Layer of your AI bill | Meta's published offer for Muse | Where credits do exist |
|---|---|---|
| Model API tokens | None | Tracked at getaiperks.com |
| Image generation | None | Tracked at getaiperks.com |
| Speech and transcription | None | Tracked at getaiperks.com |
| Cloud and GPU compute | None | Tracked at getaiperks.com |
$7.7M in credits across 194 companies, and a Muse Spark integration that took a base URL change to build takes a base URL change to move. Compare the funded option at getaiperks.com before committing to the unfunded one.
Frequently Asked Questions
How much does the Meta Model API cost?
Muse Spark 1.1, 1.2 and 1.3 all cost $1.25 per million input tokens and $4.25 per million output tokens. Muse Image is $0.01 per image and Muse Voice Transcribe is $3.00 per 1,000 minutes. A Contributor tier offers the same models at $0.10 and $0.20 per million tokens.
What is the Meta Model API Contributor tier?
Contributor is a cheaper pricing tier on the Meta Model API: $0.10 per million input tokens and $0.20 per million output tokens instead of $1.25 and $4.25. In exchange, Meta trains on your prompts and completions. It suits experimentation and output-heavy generation, not customer data or proprietary system prompts.
Does Meta offer free Muse Spark credits for startups?
No. Meta has no dedicated Muse or Meta Model API credit program. Its developer-programs page lists only VR and MR programs, unrelated to Muse. Claims of $20 in free API credits are unverified. Providers that do fund startups are tracked at getaiperks.com.
How much does the Muse app cost per month?
The Muse consumer agent has a free tier, Power at $20 per month and Maximum at $100 per month. The tiers differ by usage volume, not by features. Meta has not published the free tier's task or token limits, nor any Meta Model API allowance attached to the paid plans. Credits for API alternatives are listed at getaiperks.com.
Is muse.ai still the video hosting company?
No. muse.ai was an independent video hosting and search company founded by Antonio Roldao. It rebranded to Skiv on 3 March 2026, with an embed URL migration deadline of 10 March. Meta later took over the vacated domain as the web front end for its agent.
Is Microsoft Muse the same as Meta Muse?
No connection at all. Microsoft's Muse, announced in February 2025, is the WHAM gameplay-generation research model built with Xbox Game Studios and Ninja Theory. Meta's Muse is a 2026 consumer AI agent. Startup credit programs across AI providers are tracked at getaiperks.com.
Pay list price if you want to. Check what is already funded first.