Muse Wonder is Meta's frontier model for coding, long-running agents and computer use, introduced on October 12, 2026. On the Meta Model API it costs $4.00 per million input tokens, $0.20 per million cached input tokens and $20.00 per million output tokens, with a 1M-token context window. The stronger Muse Wonder Full has no per-token price: it comes only with a Muse Code subscription on the Power Usage ($50/month) or Maximum Usage ($200/month) plan.
Short answers for muse wonder pricing, muse wonder price, muse wonder free, muse wonder full and muse wonder subscription: there is no free tier or free credit that Meta documents for Muse Wonder; the cheapest official way in is the $5/month Muse Code Everyday Usage plan; and per token it costs about 3.2x Muse Spark 1.3 on input and 4.7x on output. Below, the muse wonder model itself is explained first, then pricing. All figures come from Meta's official pages, checked October 15, 2026 (Beijing time). This update also places Wonder inside the wider Muse lineup (Wonder vs Spark for agents).
Muse Wonder Model: Active Memory, Active Focus and Agent Tasks
Meta's launch post describes the Muse Wonder model by two traits it says matter for personal and professional agents:
- Active Memory: carrying important context about you across conversations and actively retrieving the details relevant to the task at hand.
- Active Focus: keeping at a hard goal by trying alternative approaches, working through blockers and adapting to new information, while asking for clarification when needed and checking in before important actions. Meta measures this with FocusBench and with SWE-Marathon, a test of long-running software projects.
Meta also highlights domain work in law, finance and health (including Harvey's Legal Agent Benchmark) and frontier coding. Its demos show Muse Wonder working in a browser and across apps: building a running plan from personal preferences, designing and ordering custom swag from Instagram photos, preparing a wedding seating chart from matching rules, and building a playable 3D kitchen game. Meta says both Muse Wonder and Muse Wonder Full were trained at its Prometheus data center and build on larger base models than Muse Spark 1.3. These are Meta's descriptions.
Muse Product Line: Muse AI, Muse Wonder and Muse Spark
Meta's developer docs describe Muse as a product family, not a single model. On Models - Meta Model API, Meta ships six lines: Muse Wonder, Muse Spark, Muse Image, Muse Voice Transcribe, and Segment Anything Model (SAM) are hosted on Meta Model API; Muse Glimmer is open-weight and self-hosted.
For agent and coding work, the docs put two text models side by side:
| Line | Role (Meta's wording) | Where |
|---|---|---|
Muse Wonder (muse-wonder-1.0) |
Most capable model for complex software engineering and long-horizon agents; plans, delegates to sub-agents, parallel tools, computer use | Model API pay-as-you-go (Standard only) and every Muse Code plan; Muse Wonder Full only on Power ($50/month) and Maximum ($200/month) |
Muse Spark (muse-spark-1.3 current) |
Agentic and coding work for high-volume or cost-sensitive loops; Standard and cheaper Contributor modes | Model API and Muse Code (1.3 and 1.2); Spark 1.1 is API-only |
Meta's choose-guide: pick Wonder when capability matters more than cost per token (multi-file coding, multi-hour agents); pick Spark for latency-sensitive or cheaper sub-agent work. Both take text, image, video, audio, and PDF and return text with a 1,048,576-token context window. Consumer Muse app experiences are separate: Meta's launch post says Wonder will soon power experiences in Muse and roll out first to subscribers—it does not define a separate "Muse AI" API model ID.
For a wider product map that includes Muse Code and subscriptions, see AI Rank's Muse Meta overview. Pricing for Wonder itself is in the sections below.
Muse Wonder Pricing: API Rates per Million Tokens
From Meta's Muse Wonder developer page and pricing and rate limits doc:
| Item | Muse Wonder (muse-wonder-1.0) |
Muse Wonder Full (muse-wonder-full-1.0) |
|---|---|---|
| Input | $4.00 / 1M tokens | Not sold per token |
| Cached input | $0.20 / 1M tokens | Not sold per token |
| Output | $20.00 / 1M tokens | Not sold per token |
| Context window | 1M tokens | 1M tokens |
| Where | Meta Model API (pay-as-you-go) and Muse Code | Muse Code subscription only |
A few details from Meta's pricing doc matter for budgeting. Muse Wonder runs in Standard mode only on the Model API, so there is no discounted Contributor-mode version of it. Meta charges no long-context premium: the rate is the same whether the 1M window is nearly empty or nearly full. Web search grounding adds $2.50 per 1,000 queries. The default rate limit is 3,000 requests and 4M tokens per minute per team. OpenRouter lists meta/muse-wonder at the same $4 / $20 rates.
Muse Wonder Price vs Muse Spark 1.3
Muse Spark 1.3 is Meta's cheaper model on the same API. Meta's Model API page lists both (USD per 1M tokens):
| Model | Input | Cached input | Output |
|---|---|---|---|
Muse Wonder (muse-wonder-1.0) |
$4.00 | $0.20 | $20.00 |
Muse Spark 1.3 (muse-spark-1.3, Standard) |
$1.25 | $0.15 | $4.25 |
Muse Spark 1.3 Contributor (muse-spark-1.3-contributor) |
$0.10 | $0.002 | $0.20 |
Against Standard Muse Spark 1.3, Muse Wonder costs 3.2x on input, about 4.7x on output and 1.33x on cached input. Against the Contributor variant, where Meta may use your content to train future models, the gap is 40x on input and 100x on output. Spark 1.3 also has a 1M context window, so the price difference buys capability, not a bigger window.
Muse Wonder Free Access: None Documented
Not in any form Meta documents. As of October 15, 2026, none of Meta's Muse Wonder page, Model API page, pricing doc, Muse Code subscription doc or launch post lists a free tier, free trial or free API credits for Muse Wonder. OpenRouter has no free variant of it either.
- Cheapest paid access: Muse Code's Everyday Usage plan, $5/month, which Meta says includes Muse Wonder (10–50 prompts every 5 hours). The High Usage plan is $15/month with 5x the usage.
- Muse consumer app: Meta's launch post says Muse Wonder "will soon power experiences in Muse" and will roll out first to subscribers. It does not say when, or whether free users will get it.
- Lowest per-token cost: with no free option, pay-as-you-go at $4 / $20 is the cheapest way to try it in your own code without a subscription.
Muse Wonder Full and the Subscription Plans
Meta calls Muse Wonder Full "our most advanced model yet." The Muse Code subscription doc and Muse Code page list four plans:
| Muse Code plan | Price | Muse Wonder | Muse Wonder Full | Usage vs Everyday |
|---|---|---|---|---|
| Everyday Usage | $5.00/month | Yes | No | 1x (10–50 prompts per 5 hours) |
| High Usage | $15.00/month | Yes | No | 5x |
| Power Usage | $50.00/month | Yes | Yes | 20x |
| Maximum Usage | $200.00/month | Yes | Yes | 80x |
Subscriptions work only through the Muse Code CLI; any other API key on the account is billed pay-as-you-go. Plans default to Contributor mode, which gives more usage in exchange for letting Meta train on your content, and you can switch it off in settings. So the cheapest official route to Muse Wonder Full is $50/month.
Muse Wonder on AI Rank: Arena and Usage
Muse Wonder is now on the Arena leaderboard. In the Arena text snapshot published October 13, 2026 (style control), muse-wonder-max ranks #2 overall and #2 in coding, behind Gemini 4 Argon high in both:
| Arena category | Rank | Rating | 95% CI | Votes |
|---|---|---|---|---|
| Overall | #2 | 1512.7 | 1499.4–1525.9 | 1,944 |
| Coding | #2 | 1558.0 | 1531.8–1584.1 | 487 |
| Multi-turn | #2 | 1544.4 | 1508.9–1580.0 | 256 |
| Instruction following | #5 | 1502.9 | 1480.2–1525.7 | 653 |
| Creative writing | #8 | 1485.7 | 1456.2–1515.2 | 427 |
Vote counts are still small, and Muse Wonder's overall interval overlaps both Gemini 4 Argon high (1515.9–1533.6) and Claude Opus 5.5 high (1500.5–1515.6), so #2 is not a settled position. Muse Spark 1.3 max sits at #11 overall (1494.2).
→ Track it on AI Rank: Muse Wonder — Arena rating, rank and 90-day history
On usage, meta/muse-wonder-20261012 placed #34 in AI Rank's OpenRouter daily text ranking for October 13 (60.2B tokens) and #85 for the week of October 7–13 (72B tokens, first week listed). Muse Spark 1.3 Contributor, the cheap variant, was #10 that day with 432B tokens.
→ Live usage: See today's order in the OpenRouter model usage ranking on AI Rank (tokens, daily window).
Bottom Line
To call Muse Wonder from your own product, pay $4 / $20 per million tokens on the Model API. To use it inside Muse Code, the $5 Everyday plan is the entry point, and Muse Wonder Full starts at $50/month. If cost matters more than top-end capability, Muse Spark 1.3 is a fraction of the price.
Compare before you choose: pricing changes and rankings move daily — check the current model rankings on AI Rank before you commit.
Or 📱 get the AI Rank app to follow them on your phone.
Prepared with automated editorial assistance and checked against official sources and dated AI Rank data on October 15, 2026. We did not run our own benchmarks for this article.