Muse Code pricing: a monthly plan, or pay per token.
Meta sells Muse Code two ways: monthly plans from $5, or pay-per-token billing through its Model API. Switch models mid-session with /model.
Unofficial fan site — plans and billing happen on Meta’s site.
Monthly plans
Added when Muse Code left beta on August 31, 2026; Maximum joined on October 12. Plans meter requests, not tokens, and start in Contributor mode.
- ✓10–50 requests every 5 hours, including image and video uploads
- ✓Muse Spark, Muse Wonder, and Muse Image
- ✓Voice mode and web search
- ✓Everything in Everyday Usage
- ✓5× the usage of Everyday Usage
- ✓More multimodal inputs
- ✓Everything in High Usage
- ✓20× the usage of Everyday Usage
- ✓Muse Wonder Full, Meta’s most capable model
- ✓Early access to new features, higher file uploads
- ✓Everything in Power Usage, including Muse Wonder Full
- ✓80× the usage of Everyday Usage
- ✓For all-day, multi-agent workloads
Plan details as listed in Meta’s Muse Code subscription docs, checked October 19, 2026. High and Power launched at 3× and 10×; Meta raised them to 5× and 20× in September without an announcement. Final prices appear at checkout and may vary by region.See plans on Meta →
Pay per token
No plan required: bill Muse Code through the Meta Model API instead. Two tiers of the same model — the difference is what happens to your data.
Muse Wonder on the API (muse-wonder-1.0): $4.00 input, $0.20 cached, $20.00 output per 1M tokens — Standard only, with no Contributor variant. Muse Wonder Full isn’t sold per token at all; it comes only with the Power and Maximum plans.
Estimate your pay-per-token cost
A heavy day of agentic coding uses roughly 20–60M input and 2–6M output tokens.
Plans meter requests instead: Everyday Usage covers roughly 10–50 requests every 5 hours. If your Standard estimate lands well above $15 a month, compare it with a plan.
Billing questions
Which plan should I pick?
Everyday Usage ($5) fits a few sessions a day — roughly 10–50 requests every 5 hours. High Usage ($15) gives 5× that. Power Usage ($50) gives 20× and is the cheapest plan with Muse Wonder Full; Maximum Usage ($200) gives 80× for all-day, multi-agent work. If you drive Muse Code from CI or scripts, an API key with pay-per-token billing is metered by tokens instead of requests.
Subscription or pay-per-token: which is cheaper?
It depends on volume. A plan is a flat monthly fee with a request allowance; pay-per-token bills every token at Model API rates. Light use on the Contributor tier can cost cents a day, while a heavy month on Standard can reach hundreds of dollars — that’s where a flat plan pays off. Plug your numbers into the estimator above.
Is Muse Code free?
No. The CLI is free to install, but using it takes a monthly plan (from $5) or pay-per-token billing, and Meta hasn’t announced a free tier. The cheapest route is the Contributor tier — $0.10 per 1M input tokens — in exchange for letting Meta use your sessions. Muse Glimmer, Meta’s open-weight model, is free to run locally, but it is a separate model, not Muse Code.
What does the Contributor tier share?
Data from muse-spark-1.3-contributor may be used to improve Meta’s products. The standard muse-spark-1.3 model is never used to improve Meta’s products.
Is there a Contributor tier for Muse Spark 1.3?
Yes. muse-spark-1.3-contributor keeps the 1.2 Contributor rates — $0.10 per 1M input tokens, $0.002 cached, $0.20 output — and is also listed on OpenRouter.
Can I switch mid-session?
Yes. /model switches models mid-session — for example from muse-spark-1.3 to muse-wonder-1.0 — and usage bills at whichever model served the tokens.
Do cached or repeated context tokens count?
Repository context is cached across a session. Cached input tokens bill at a reduced rate — $0.002 / 1M on Contributor, $0.15 / 1M on Standard — instead of the full input price.
Can I get zero data retention?
Meta is beginning to accept requests for zero data retention. Contact Meta sales to request it for your organization.
Is a Meta Muse subscription the same as a Muse Code plan?
No. Meta Muse, launched September 8, 2026, is a consumer personal agent with its own plans: free, Power at $20 a month, and Maximum at $100. Muse Code plans — Everyday $5, High $15, Power $50, Maximum $200 — are for the coding agent. Both products have a tier called Power and one called Maximum, but they are different products, priced separately.
Which plans include Muse Wonder Full?
Only Power Usage ($50) and Maximum Usage ($200). Everyday and High include Muse Wonder, the API model, but not Wonder Full — and Wonder Full isn’t available on the Model API at any price, so a plan is the only way to use it.
What happens when I hit my plan limit?
You can upgrade, which takes effect immediately with a prorated charge, or wait for the limit to refresh; Maximum users can only wait. There are no paid top-ups. To keep working, add a Model API key and pay per token — but API keys can’t reach Wonder Full.
Do Muse Code plans use my code for training?
By default, yes: subscriptions start in Contributor mode, which trades your prompts and code for more plan usage. Turn it off in /settings and future content isn’t used for training, but your allowance shrinks. Don’t send confidential code while Contributor mode is on.
Can I change or cancel my plan?
Upgrades apply immediately and are prorated. Downgrades start at the next billing cycle, and you keep your current limit until then. If you cancel, access lasts to the end of the cycle — cancel at least 24 hours before the billing date to avoid the next charge.
Where do I manage, change, or cancel my Muse Code subscription?
In Meta’s Accounts Center (accountscenter.meta.com): open Subscriptions, select Muse Code, then Manage your subscription. Typing /upgrade in the CLI opens the same page in your browser. Payment methods and billing history live there too. Cancelled subscriptions aren’t refunded unless the law requires it, and uninstalling the CLI doesn’t cancel anything.
Is there a weekly limit?
Meta describes plan usage in 5-hour windows — Everyday Usage covers roughly 10–50 prompts every 5 hours, and the higher plans multiply that — and its subscription docs don’t mention a separate weekly cap. Since CLI 1.4.13, muse usage shows how much of your allowance is left. Pay-per-token billing has per-minute rate limits instead (3,000 requests a minute on Muse Spark 1.3 and Muse Wonder, 150 on the Contributor model).
Is there a free trial or a student discount?
Meta hasn’t announced either for Muse Code: its subscription docs list no trial and no student or education pricing. The cheapest ways in are the $5 Everyday Usage plan or pay-per-token on the Contributor model, where light use costs cents a day.
What is a token, and how many does a session use?
A token is a small chunk of text — in English, roughly four characters. Input tokens cover your prompt, system instructions, and the conversation so far; output tokens are what the model writes, including its reasoning, and cost more. Code and non-English text use more tokens per word. A heavy day of agentic coding runs roughly 20–60M input and 2–6M output tokens. Plans count prompts rather than tokens, so tokens only matter for pay-per-token billing.
Do Muse invite codes work for Muse Code?
No. The invite codes people share, which give bonus tokens to both sides, belong to Meta Muse, the consumer agent at muse.ai. Muse Code has no invite or referral program, and those tokens don’t pay for Muse Code plans or Model API usage.