Muse Wonder Pricing: $4/$20 per 1M Tokens

What Muse Wonder Actually Costs: The $4/$20 Token-Price Signal

Muse Wonder is reported to cost $4 per million input tokens, $20 per million output tokens, and $0.20 per million cached input tokens through the Meta Model API. That rate comes from several accounts on launch day, October 12, 2026, and was not listed in Meta's own announcement — treat it as widely reported but not officially confirmed. At those numbers, Muse Wonder lands exactly on Claude Opus 5.5's published price and roughly 60% below Claude Fable 5.1.

Key Takeaways

• Reported API price: $4/M input, $20/M output, $0.20/M cached input — attributed to multiple community posts, first surfaced by @LuminaBench and echoed by others.
• Not in the official post: Meta's launch blog confirmed availability but did not publish per-token rates, so the $4/$20 figure stays unconfirmed as of October 13, 2026.
• Wonder Full has no API price: the more capable variant is restricted to Muse Code subscribers, so there is no per-token rate to quote.
• Matches Opus 5.5, undercuts Fable 5.1: the same $4/$20 as Opus 5.5; about 60% cheaper than Fable 5.1's cited $10/$50.
• 1M-token context is reported for both variants, which makes the $0.20 cached-input rate the number that matters for long agent runs.
• Early limits were buggy: Meta said it fixed a quota-drain issue the same day, so subscription usage caps are still settling.

Muse Wonder Pricing Breakdown

The table below gathers every price point tied to the Muse Wonder family. Confirmed availability comes from Meta's launch material; the per-token figures are community-reported and marked as such.

Item	Reported rate	Status	Source basis
Muse Wonder — input	$4.00 / 1M tokens	Reported, not official	Launch-day community posts
Muse Wonder — output	$20.00 / 1M tokens	Reported, not official	Launch-day community posts
Muse Wonder — cached input	$0.20 / 1M tokens	Reported, not official	Launch-day community posts
Muse Wonder Full — API	Not yet confirmed	Subscriber-only, not on API	Meta launch material
Muse Code subscription	"Maximum" plan reported at $100/mo, ~3B weekly tokens	Single-source, unconfirmed	One community report
Context window	1,000,000 tokens	Reported	Multiple community posts
Free API tier	Not yet confirmed	—	—

One early post listed cached input at $0.02 rather than $0.20; the $0.20 figure is the one repeated across the most accounts, so it is the safer reference until Meta publishes a rate card.

Source: @LuminaBench

What That Costs in Practice

All figures below use the reported $4/M input and $20/M output rate. Each row shows the arithmetic so you can re-run it if Meta's official numbers land differently.

Scenario	Input	Output	Math	Est. cost
Single short call	10K tokens	2K tokens	(0.010 × $4) + (0.002 × $20)	$0.08
Document Q&A	50K tokens	5K tokens	(0.050 × $4) + (0.005 × $20)	$0.30
Long agent task	100K tokens	50K tokens	(0.100 × $4) + (0.050 × $20)	$1.40
Heavy coding run	400K tokens	120K tokens	(0.400 × $4) + (0.120 × $20)	$4.00

The output rate dominates. Because Muse Wonder bills output at 5× its input rate, agent workloads that generate long plans, code, or tool-call chains spend most of their budget on the output side. The $0.20 cached-input rate is the lever for repeat context: a 500K-token system prompt reused across calls costs $0.10 per call cached versus $2.00 uncached.

A launch-day frontend test reported a Muse Wonder Full run that took 10 minutes and cost $10, against a comparable Fable 5.1 run at 30 minutes and $20 — one task, not a benchmark, but a useful order-of-magnitude anchor.

How Muse Wonder Pricing Compares

Muse Wonder's reported rate is the center of the pricing story because it mirrors one rival and badly undercuts another.

Model	Input / 1M	Output / 1M	Note
Muse Wonder (reported)	$4.00	$20.00	Community-reported
Claude Opus 5.5	$4.00	$20.00	Same on paper
Claude Fable 5.1	$10.00	$50.00	~60% pricier both sides
Muse Spark (prior gen)	$1.25	~$4.25	Meta's cheaper predecessor

As @kimmonismus put it, "Claude Fable 5 and 5.1 cost $10 per million input tokens and $50 per million output tokens, making Muse Wonder's reported $4/$20 pricing 60% cheaper for both." Against Opus 5.5 the list prices are identical, so any real cost edge has to come from token efficiency rather than the rate card. Community posts claimed Muse Wonder finished some tasks using fewer tokens, and one cost comparison claimed a roughly 2.6× lower effective price on a shared 3D prompt — again, single-task figures, not list prices. Our deeper read on that matchup lives in this Muse Wonder versus Opus 5.5 comparison.

Source: @kimmonismus

Free Tier, Limits, and Access

Muse Wonder is available today through the Meta Model API and inside Muse Code, per Meta's research blog. Muse Wonder Full is restricted to Muse Code subscribers and is not on the API, which is why it carries no per-token price. The consumer Muse app is slated to get Muse Wonder "soon," rolling out to subscribers first — no date was given.

There is no confirmed free API tier. On the subscription side, one report described a $100/month Muse "Maximum" plan with roughly 3 billion weekly tokens for the latest models, but that is a single-source claim and the model attribution is provisional.

Usage limits were rocky at launch. Several testers reported quota draining far faster than expected; Meta's Alexandr Wang acknowledged a limits bug and later said the team "identified and fixed the issue that was causing the unusually fast quota usage," adding a Muse Code reset. If you are budgeting against subscription caps rather than API tokens, expect the effective allowance to keep shifting while that settles.

For teams comparing like-priced frontier options, the identically rated Claude Opus 5.5 is the natural reference point on access and tooling. Our standalone Opus 5.5 pricing read covers that $4/$20 signal in detail.

What We Don't Know Yet

• Whether $4/$20/$0.20 is final and official. The figures are consistent across accounts but absent from Meta's own post.
• Muse Wonder Full's API price, if it ever reaches the API at all.
• Which variant the 1M context and multimodal inputs apply to, and at what tier.
• Consumer Muse pricing and any free allowance once Wonder reaches the app.
• Steady-state subscription limits after the quota-bug fix.

Frequently Asked Questions

How much does the Muse Wonder API cost? Muse Wonder is reported to cost $4 per million input tokens, $20 per million output tokens, and $0.20 per million cached input tokens. These figures come from multiple community posts on launch day and were not stated in Meta's official announcement.

Is Muse Wonder free? Muse Wonder is not documented as free at the API level. It is billed per token through the Meta Model API and bundled into Muse Code subscriptions. A free consumer tier inside the Muse app has not been confirmed.

Is Muse Wonder cheaper than Claude Opus 5.5? Muse Wonder's reported $4/$20 per-million-token rate matches Claude Opus 5.5's published rate, so on paper they are priced the same. Community cost tests claimed Muse Wonder came out cheaper on specific tasks, but those are single-task comparisons, not list-price differences.

Is Muse Wonder cheaper than Claude Fable 5.1? Yes, by the reported numbers. Fable 5 and 5.1 are cited at $10 per million input tokens and $50 per million output tokens, making Muse Wonder's reported $4/$20 roughly 60% cheaper on both input and output.

How much does Muse Wonder Full cost? Muse Wonder Full has no published API price. Meta's launch material places Wonder Full inside Muse Code subscriptions only, not on the Meta Model API, so there is no per-token rate to quote yet.

What is the Muse Wonder context window and does it change the price? Muse Wonder is reported to support a 1 million token context window. Pricing is per token, so filling a large context raises the per-call cost linearly; the $0.20 cached-input rate is designed to cut the cost of reused context.

Where can I use Muse Wonder? Muse Wonder is available through the Meta Model API and in Muse Code as of October 12, 2026. Muse Wonder Full is limited to Muse Code subscribers, and both are slated to reach the consumer Muse app for subscribers first.

What to Watch Next

Three signals will decide whether the $4/$20 number holds. First, an official Meta rate card or pricing page — until then the figure stays community-reported. Second, whether Muse Wonder Full gets an API endpoint and its own price, which would change the competitive math against Opus 5.5. Third, the post-fix subscription limits in Muse Code, since the launch-day quota bug means current caps are not representative. This page will be amended as each lands.

Try related models on Kie.ai: Claude Opus 5.5, GPT-6 Astra and Claude Fable 5.1 (Coming Soon).