Muse Wonder price on Meta’s Model API is $4.00 per million input tokens and $20.00 per million output tokens. Cached input is $0.20. That is Standard mode only. Contributor mode is not available for Wonder on the Model API.
Meta announced Muse Wonder and Muse Wonder Full on October 12 as its “most capable models yet” for personal and professional agent work, coding, memory and focus. The models are live today. The open weights are not. Meta says Wonder is too large for many people to run, and cites safety concerns. That gap — a priced API with closed weights, and a Full tier that has no public per-token price — is what people are searching for.
Muse Wonder price on Meta’s Model API
Meta’s pricing and rate-limits page, which we checked on October 13, 2026, lists Muse Wonder (muse-wonder-1.0) like this:
| Per 1M tokens (Model API) | Muse Wonder Standard | Muse Spark 1.3 Standard (same page) |
|---|---|---|
| Input | $4.00 | $1.25 |
| Cached input | $0.20 | $0.15 |
| Output | $20.00 | $4.25 |
Against Spark 1.3 Standard on the same sheet, Wonder is about 3.2× on input and about 4.7× on output. Context window for Wonder is 1 million tokens. Rate limits are 3,000 requests per minute and 4,000,000 tokens per minute. Web search grounding is $2.50 per 1,000 queries.
Muse Wonder Full has no public token price
Muse Wonder Full does not show a public per-token API rate. Meta’s model page says access is on Muse Code Power Usage ($50/month) and Maximum Usage ($200/month) plans only. The research blog says Wonder Full is available today for Muse Code subscribers. Meta has not published a Full token price.
Weights stay closed; Spark open weights are “coming soon”
There is no Hugging Face download for Muse Wonder. In an @AIatMeta reply on October 12, Meta wrote that open weights for Muse Spark are “coming soon,” and that Muse Wonder models “are large models that are too big for many people to use, in addition to potential safety concerns.” Treat the Spark open-weights line as a company plan, not a shipped file. Wonder stays closed-weight on Meta’s own statement.
What Meta names as features, and what Meta’s table claims
Meta’s research post says Muse Wonder brings together Active Memory and Active Focus for personal and professional agents, coding, memory and focus. Meta’s model page publishes its own comparison table of Muse Wonder Full and Muse Wonder against Claude Fable 5.1, GPT-6 Astra and Claude Opus 5.5. In that Meta table, Artificial Analysis Index is 56 for Full and 52 for Wonder; DeepSWE 1.1 is 79.0 for Full and 78.8 for Wonder; SWE-Marathon is 57.5 (xhigh) for Full and 53.1 for Wonder. Those are Meta’s numbers and Meta’s framing, not independent confirmations.
The catches before you budget around it
Three of them. First, Wonder on the Model API is Standard only — there is no Contributor discount on that path. Second, Full has no public per-token price; it rides Muse Code Power and Maximum Usage plans. Third, weights are closed for Wonder, and Spark’s open-weights promise is still a “coming soon.” Until Meta publishes a Full token rate or ships Spark weights, Wonder is a closed, Standard-priced API model with a Code-only Full tier on top.
Frequently asked questions
How much does Muse Wonder cost?
On Meta’s pricing page as of October 13, 2026, Muse Wonder Standard on the Model API is $4.00 per million input tokens, $0.20 for cached input and $20.00 for output. Context is 1 million tokens. Contributor mode is not available for Wonder on the Model API.
How much does Muse Wonder Full cost?
Meta has not published a public per-token API price for Muse Wonder Full. The model page says access is included with Muse Code Power Usage ($50/month) and Maximum Usage ($200/month) plans.
Are Muse Wonder weights open or downloadable?
No. Meta said on October 12 that Muse Wonder models are closed, citing size and safety concerns. Open weights for Muse Spark are described as “coming soon”; they were not a downloadable Wonder release when we checked.
How does Muse Wonder price compare to Muse Spark?
On the same Meta pricing page, Muse Spark 1.3 Standard is $1.25 input, $0.15 cached and $4.25 output per million tokens. Wonder Standard is about 3.2× input and about 4.7× output versus that Spark Standard line.
Sources
All accessed October 13, 2026.
- Meta research blog, “Introducing Muse Wonder” (Oct 12, 2026): research.meta.ai
- Meta, Muse Wonder model page (access, plans, Meta comparison table; checked Oct 13, 2026): dev.meta.ai/models/muse-wonder
- Meta, pricing and rate limits (Wonder $4/$0.20/$20, Spark 1.3 Standard, web search $2.50/1k queries, 3,000 RPM / 4,000,000 TPM; checked Oct 13, 2026): dev.meta.ai/docs/pricing-rate-limits
- Meta AI on X, Muse Wonder launch (Oct 12, 2026, about 1:02 PM ET): x.com/AIatMeta
- Alexandr Wang on X, Muse Wonder launch post (Oct 12, 2026): x.com/alexandr_wang
Related: Mistral Large 4 costs $0.68/$2.09 on sale; open weights not out yet · GPT-6.1 Sol Ultrafast costs more than Astra at standard speed · More LLM news
Last updated: October 13, 2026. To report an error, see our corrections page. Articles are drafted with AI assistance and reviewed and edited by an editor; see our editorial policy.
