# Muse Wonder Pricing and Subscription: App, Muse Code or API

> Muse Wonder has three price lists: $20 or $100 a month in the Muse app, $5–$200 in Muse Code (Wonder Full from $50) and $4/$20 per 1M tokens on Meta's API.

- URL: https://blog.laozhang.ai/en/posts/muse-wonder-pricing
- Published: 2026-11-06
- Updated: 2026-11-06
- Author: LaoZhang AI Team (https://blog.laozhang.ai/en/about)
- Topic: API Pricing & Plans
- Tags: Muse Wonder, Meta, Muse Code, Pricing, AI Subscription

---
There is no single Muse Wonder subscription. As of November 6, 2026, Meta sells access to Muse Wonder through three separate products, each with its own price list. The **Muse app**, Meta's consumer AI agent, is free on Muse Spark, and Wonder sits in its paid tier: the Power plan at $20 a month or the Maximum plan at $100. **Muse Code**, Meta's coding agent for the terminal, costs $5, $15, $50 or $200 a month; every plan includes Wonder, but Muse Wonder Full only comes with the $50 Power Usage and $200 Maximum Usage plans. The **Meta Model API** has no subscription at all: `muse-wonder-1.0` is billed at $4 per million input tokens, $0.20 per million cached input tokens and $20 per million output tokens.

So pick the product by the job. If you want Wonder for everyday questions and tasks on your phone or at muse.ai, the Muse app's $20 Power plan is the entry point. If you write code and want the stronger Wonder Full, the $50 Muse Code Power Usage plan is the cheapest way to get it, because Wonder Full has no API. If you are building Wonder into your own app or scripts, use the API and pay per token.

Two things can change that answer before you pay. Meta has launched the Muse app only in the US, Canada, Japan and Singapore, and Meta's model page says Wonder isn't currently available in South Korea on the API or in Muse Code. A Muse Code subscription also starts in Contributor mode, which lets Meta train on your code until you switch it off.

## Muse Wonder prices in the Muse app, Muse Code and the API

The three price lists don't line up because they measure different things: the Muse app meters "Muse tokens" per week, Muse Code meters usage in multiples of its cheapest plan, and the API bills real tokens. The table below combines Meta's [Muse subscription help page](https://www.meta.com/help/subscriptions/1021145227643680/), the [Muse Code product page](https://dev.meta.ai/products/muse-code), the [Meta Model API pricing and rate limits page](https://dev.meta.ai/docs/pricing-rate-limits) and the [Meta models page](https://dev.meta.ai/docs/models), as of November 6, 2026. All prices are in US dollars per month unless marked otherwise.

| Product and plan | Price | Muse Wonder | Muse Wonder Full | How usage is measured |
| --- | --- | --- | --- | --- |
| Muse app, Free | $0 | No, the free tier runs Muse Spark | No | A free usage limit; Meta publishes no number |
| Muse app, Power plan | $20 | Paid tier; users report Power switches to Wonder | No | 500M Muse tokens per week |
| Muse app, Maximum plan | $100 | Paid tier; Meta names no plan | No | 3B Muse tokens per week |
| Muse Code, Everyday Usage | $5 | Yes | No | "10-50 prompts every 5 hours" |
| Muse Code, High Usage | $15 | Yes | No | 5x Everyday |
| Muse Code, Power Usage | $50 | Yes | Yes | 20x Everyday |
| Muse Code, Maximum Usage | $200 | Yes | Yes | 80x Everyday |
| Meta Model API, pay-as-you-go | No monthly fee | Yes, as `muse-wonder-1.0` | No | $4 input, $0.20 cached input, $20 output per 1M tokens |

So a Muse subscription costs $20 or $100 a month, and those are the only two paid Muse app plans. Watch the names: "Power" and "Maximum" exist in both the Muse app ($20 and $100) and Muse Code ($50 and $200), and they are different plans with different prices, bought in different places. Muse tokens are Meta's own allowance unit for the app, and Meta publishes no conversion to API tokens or dollars.

![Muse app, Muse Code and Meta Model API plans side by side, showing which include Wonder or Wonder Full and that Power and Maximum cost $20 and $100 in the app but $50 and $200 in Muse Code](https://blog.laozhang.ai/posts/en/muse-wonder-pricing/img/muse-wonder-plans-by-product.webp)

Inside Muse Code, the bigger plans give more usage per dollar. High Usage gives 5 times Everyday's allowance for 3 times the price, while Power Usage (20x for $50) and Maximum Usage (80x for $200) both work out to twice Everyday's usage per dollar. Meta doesn't explain the "10-50 prompts" range further, so the base those multiples apply to isn't fixed either.

The API side has a few extra charges and limits. Web search grounding adds $2.50 per 1,000 search queries on top of tokens, and there is no long-context surcharge across Wonder's 1,048,576-token context. Meta lists no free tier and no batch discount for Wonder, and the cheaper Contributor mode that Muse Spark offers on the API is "Not available" for Muse Wonder. Rate limits are 3,000 requests and 4,000,000 tokens per minute per team.

## Which one to pay for: Muse app, Muse Code or the API

Choose by where you will use Wonder, because the three products don't substitute for each other: a Muse app plan gives no API access, and a Muse Code plan doesn't work outside the Muse Code CLI.

If you want a personal agent for questions, planning and everyday tasks, the Muse app is the product, and the Power plan is the cheapest paid plan in it. Move to the Maximum plan only if you run out of the 500M weekly Muse tokens, because it gives six times the weekly allowance for five times the price. If you are still deciding between Meta's agent and its competitors, [ChatGPT Dots vs Meta Muse vs Grok Bot vs Claude Duet: Entry Costs](https://blog.laozhang.ai/en/posts/chatgpt-dots-vs-meta-muse-vs-grok-bot) compares what each costs to start.

If you code in a terminal, Muse Code is the product, and you can start low. Wonder is on every plan, including the $5 Everyday Usage plan, and an upgrade takes effect immediately with a prorated charge for the rest of the billing cycle. Go to the $50 Power Usage plan when you either hit the lower plans' limits or want Muse Wonder Full, which Meta describes as its higher-capability variant and which no other product sells. Muse Code can also run on a pay-as-you-go API key instead of a subscription, billed at the API prices above.

If you want Wonder inside your own product, in scripts, or in another coding tool, pay per token on the Meta Model API. The key a Muse Code subscription creates only works in Muse Code, so per-token billing is the only way to use Wonder anywhere else. On the API, Wonder runs in Standard mode only, which means your requests are not used for training. Getting a key and making the first request is a separate setup task, covered in [Meta Muse Wonder API: Setup, Pricing and Cost per Request](https://blog.laozhang.ai/en/posts/muse-wonder-api). If you're weighing Wonder against Anthropic's top model for the same API budget, [Muse Wonder vs Opus 5.5: Benchmarks and Cost at the Same $4/$20](https://blog.laozhang.ai/en/posts/muse-wonder-vs-opus-5-5) compares them per task.

## Is the $50 Muse Code plan worth it compared with paying per token?

At list prices, $50 buys about 7 of Artificial Analysis's long agent tasks on Muse Wonder at max effort, or about 11 at high effort, on the API. If you'd run fewer jobs of that size in a month, paying per token costs less than the $50 plan; if you'd run more, the flat plan starts to pay off, and it is also the only way to get Wonder Full. These figures are a calculation made on November 6, 2026 from Meta's list prices and the costs Artificial Analysis published for its Intelligence Index v5.0.1, not a measurement of Wonder's usage.

The method is a simple division. Artificial Analysis reports that Muse Wonder costs $6.95 per index task at max effort and $4.56 per task at high effort ([Muse Wonder vs Claude Opus 5.5 on Artificial Analysis](https://artificialanalysis.ai/models/comparisons/muse-wonder-vs-claude-opus-5-5), [Muse Wonder (high) on Artificial Analysis](https://artificialanalysis.ai/models/muse-wonder-high)). Dividing each monthly fee by those costs shows how many such tasks the same money buys on pay-as-you-go:

| Plan | Monthly fee | Tasks at max effort | Tasks at high effort |
| --- | --- | --- | --- |
| Muse Code Everyday Usage | $5 | 0.7 | 1.1 |
| Muse Code High Usage | $15 | 2.2 | 3.3 |
| Muse app Power plan | $20 | 2.9 | 4.4 |
| Muse Code Power Usage | $50 | 7.2 | 11.0 |
| Muse app Maximum plan | $100 | 14.4 | 21.9 |
| Muse Code Maximum Usage | $200 | 28.8 | 43.9 |

Read the table with its limits in mind. An Artificial Analysis task is a long agentic benchmark job, averaging 96.1k output tokens and 12.4 minutes at max effort, not a chat message. Meta doesn't publish how many tasks of that size any plan allows, so the table shows what each fee would buy on the API, not what the plan itself lets you run. The Muse app rows are there only for scale, since app plans give app usage and no API access. The calculation also assumes Meta charges nothing for writing to the cache, as its pricing page lists no such fee.

The same calculation shows how much caching matters if you choose the API. Artificial Analysis spent $5,414 running the whole index on Wonder at max effort, with 11.4B input tokens and 113M output tokens. Output accounts for $2,260 of that (113M × $20 per million), so input cost $3,154, which is only possible if about 98.0% of input tokens were billed at the $0.20 cached rate (solving 11,400M × ($0.20 × share + $4 × (1 − share)) = $3,154). With no cache hits, the same tokens would have cost $47,860, or 8.8 times as much.

For a single request, the cost is cached input × $0.20 + uncached input × $4 + output × $20, all per million tokens. A coding-agent turn with 200k input tokens, 90% of them cached, and 8k output tokens costs $0.276 ($0.036 + $0.08 + $0.16). The same turn with nothing cached costs $0.960. A smaller request with 50k input tokens, half cached, and 4k output costs $0.185. At $0.276 a turn, $50 pays for about 180 cached turns of that size.

![Bar chart of the Muse Wonder API bill for a full benchmark run, $5,414 with about 98% of input cached versus $47,860 with no cache hits, plus one coding-agent turn at $0.276 cached versus $0.960 uncached](https://blog.laozhang.ai/posts/en/muse-wonder-pricing/img/muse-wonder-api-cache-cost.webp)

What $50 adds that money can't buy on the API is Muse Wonder Full. Artificial Analysis scores it 50 on Intelligence Index v5.0.1, against 44 for Wonder at max effort, but lists no cost for it because there is no public API price ([Muse Wonder Full on Artificial Analysis](https://artificialanalysis.ai/models/muse-wonder-full)). If that gap matters for your work, the $50 plan is the entry price regardless of the per-task math.

## Which Muse app plan runs Muse Wonder: Power or Maximum?

Muse Wonder is not free in the Muse app: the free tier runs Muse Spark, and Wonder is in the paid tier. Mark Zuckerberg said on Meta's [Q3 2026 earnings call on October 21](https://www.fool.com/earnings/call-transcripts/2026/10/21/meta-meta-q3-2026-earnings-call-transcript/) that Spark "is what is currently the default across the free tier" and that people "can access" Wonder "in the paid tier today". The API doesn't make it free either, because Meta lists no free tier for `muse-wonder-1.0`.

What Meta hasn't said is whether both paid app plans carry Wonder. The help page lists prices and Muse token allowances but names no model, and "paid tier" doesn't separate Power from Maximum. CNBC reported on October 12, 2026 that Muse app users need a $20 or $100 monthly plan to access the model. Users in Japan reported on X between October 26 and November 4 that upgrading to the Power plan switched the app from Spark to Wonder automatically, and that replies got slower afterwards ([one such report](https://x.com/AIVIS_GIGER/status/2117645098146312531)). After you upgrade, check which model the app shows before you rely on it.

The free plan's allowance isn't published either: Meta's help page says only that Muse "is available for free with a usage limit". Zuckerberg also said Meta is working on "a more efficient version" of Wonder to "roll out more broadly", without a date, so the free tier may change.

## What you give up with a Muse Code subscription

A Muse Code subscription comes with four conditions that the API doesn't have, all from Meta's [Muse Code subscriptions page](https://dev.meta.ai/docs/muse-code/subscriptions) as of November 6, 2026. Check them before choosing it over pay-as-you-go.

**Meta trains on your code by default.** "Muse Code subscriptions default to Contributor mode", which gives you more usage in exchange for letting Meta use your content to train future models; Meta doesn't say how much more. To turn it off, type `/settings` in Muse Code, select the Contributor mode row under Model, and press Enter. The setting is saved per computer, and the first interactive session on a new computer turns it on and shows a notice, so repeat it on every machine. For a single run or a CI job, pass `--contributor-mode off` or set `MUSE_CONTRIBUTOR_MODE=0`; the flag wins if both are set. Turning it off covers future content only. Wonder and Wonder Full are available with Contributor mode on or off.

**The plan only works in the Muse Code CLI.** The subscription applies to the API key created during Muse Code setup, which is "for use with Muse Code only". Any other key you create on the same account is billed pay-as-you-go, so you can't point another editor or your own app at your plan.

**There are no refunds.** An upgrade is immediate and prorated, a downgrade takes effect at the next billing cycle, and a cancellation keeps access until the end of the cycle. Cancel at least 24 hours before the next billing date to avoid the next charge, because Meta gives no credit or refund for canceled subscriptions unless the law requires it.

**Maximum Usage has a ceiling.** When you hit a plan's limit you can upgrade or wait for it to refresh, but on the $200 Maximum Usage plan the only option is to wait.

## Where you can buy Muse Wonder: countries and South Korea

Where you can pay depends on the product. The Muse app launched in the US (September 8, 2026), Canada (September 18), Japan (October 23) and Singapore (October 27), and Meta's help page says Muse subscriptions are "in limited testing and aren't available in all locations yet". Meta hasn't announced the app in the UK, the EU, South Korea, Russia, mainland China, Hong Kong or Taiwan.

Muse Code and the Meta Model API aren't tied to the app's launch list, but Meta doesn't publish a full list of where they work either. Its [Geographic Use Policy](https://dev.meta.ai/legal/geographic-use-policy), last updated October 12, 2026, says the services "are not available in all jurisdictions" and may be used "only in jurisdictions where Meta has enabled access". The places Meta does name are the policy's Restricted Territories, which include mainland China (with Hong Kong and Macau), Russia, Belarus, Iran, North Korea, Cuba, Syria, Venezuela, Pakistan, Afghanistan and Cambodia, among others, and, on its models page, South Korea for Wonder (below). So a reader in the UK or Spain, for example, can't subscribe to the Muse app yet and isn't on the restricted list, but should confirm Muse Code or API access at sign-up and checkout, because Meta says subscription benefits and availability "may vary by region". The same policy forbids using "technical measures to circumvent" Meta's geographic controls, and moving to a restricted territory revokes a Muse Code subscription immediately.

South Korea is a special case. Meta's [models page](https://dev.meta.ai/docs/models) says "Muse Wonder and Muse Wonder Full aren't currently available in South Korea, on Model API or in Muse Code", with no reason or date, even though the Geographic Use Policy doesn't list Korea.

## Muse Wonder pricing: other questions

### Does setting reasoning effort to max give me Muse Wonder Full?

No. Meta's models page says Muse Wonder Full "is available only in Muse Code with a subscription, so setting reasoning effort to "max" on Muse Wonder doesn't select it." On the API, max is just the highest of Wonder's effort levels, which run from minimal to max. Wonder Full requires the Muse Code Power Usage or Maximum Usage plan.

### Is Muse Wonder cheaper on OpenRouter?

No. As of November 6, 2026, OpenRouter lists `meta/muse-wonder` at the same $4 input, $0.20 cached input and $20 output per million tokens as Meta, with web search at $0.0025 per search, the same as Meta's $2.50 per 1,000. OpenRouter has no Muse Wonder Full listing.

### Can I pay for the Muse app in yen or Singapore dollars?

Meta publishes Muse app prices only in US dollars: its help page lists both paid plans in dollars, and Meta's Japan and Singapore launch posts give no local price. As of November 6, 2026, the App Store web listings for Muse in the US, Japan, Canada, Singapore and Mexico show that the app has in-app purchases but don't list their prices.

### Can Muse Wonder make images or videos?

No. Muse Wonder accepts text, images, video, audio and PDFs as input but only returns text. For video, see [Muse Video API: No Meta Endpoint Yet, How to Call It in Python](https://blog.laozhang.ai/en/posts/muse-video-api).

## Sources

External pages this guide links to, in the order they appear. Last updated 2026-11-06.

- [Muse subscription help page](https://www.meta.com/help/subscriptions/1021145227643680/) (meta.com)
- [Muse Code product page](https://dev.meta.ai/products/muse-code) (dev.meta.ai)
- [Meta Model API pricing and rate limits page](https://dev.meta.ai/docs/pricing-rate-limits) (dev.meta.ai)
- [Meta models page](https://dev.meta.ai/docs/models) (dev.meta.ai)
- [Muse Wonder vs Claude Opus 5.5 on Artificial Analysis](https://artificialanalysis.ai/models/comparisons/muse-wonder-vs-claude-opus-5-5) (artificialanalysis.ai)
- [Muse Wonder (high) on Artificial Analysis](https://artificialanalysis.ai/models/muse-wonder-high) (artificialanalysis.ai)
- [Muse Wonder Full on Artificial Analysis](https://artificialanalysis.ai/models/muse-wonder-full) (artificialanalysis.ai)
- [Q3 2026 earnings call on October 21](https://www.fool.com/earnings/call-transcripts/2026/10/21/meta-meta-q3-2026-earnings-call-transcript/) (fool.com)
- [one such report](https://x.com/AIVIS_GIGER/status/2117645098146312531) (x.com)
- [Muse Code subscriptions page](https://dev.meta.ai/docs/muse-code/subscriptions) (dev.meta.ai)
- [Geographic Use Policy](https://dev.meta.ai/legal/geographic-use-policy) (dev.meta.ai)