By the toolist.ai editors. Researched from public sources, not hands-on tested by toolist.ai. Updated October 13, 2026; prices checked October 13, 2026.
Muse Wonder: price, context window and benchmarks
Muse Wonder is Meta's new reasoning model for coding and long-running agents, at $4 per million input tokens and $20 per million output tokens on the Meta Model API, against $1.25 and $4.25 for Muse Spark 1.3. Artificial Analysis puts it at 52.4 on its Intelligence Index at max effort. Muse Wonder Full, which Meta calls the higher-capability variant, is only in Muse Code, on the $50 and $200 monthly plans.
Best for
- Long-running coding and agent work where, in Meta's words, capability matters more than cost per token
- Muse Code users, since every subscription plan includes it, from $5 a month
- Long prompts and documents, since cached input costs $0.20 per million tokens and there is no long-context surcharge
Key facts
| Fact | Muse Wonder | Source |
|---|---|---|
| Provider | Meta | Meta Model API docs: Pricing and rate limits |
| API model name | muse-wonder-1.0 | Meta Model API docs: Models |
| Released | October 12, 2026 | Meta AI: Muse Wonder: A Frontier Model For Personal & Professional AI Agents |
| Input price | $4 per million tokens | Meta Model API docs: Pricing and rate limits |
| Cached input price | $0.20 per million tokens | Meta Model API docs: Pricing and rate limits |
| Output price | $20 per million tokens | Meta Model API docs: Pricing and rate limits |
| Context window | 1,048,576 tokens | Meta Model API docs: Models |
| Input and output | Text, images, video, audio and PDF in; text out | Meta Model API docs: Models |
| Reasoning | Always reasons; reasoning cannot be turned off. Effort levels minimal to max; when no effort is set the model picks its own level. Max on the API does not select Muse Wonder Full | Meta Model API docs: Reasoning |
| API and cloud platforms | In preview on the Meta Model API (Responses, Chat Completions and Messages endpoints; Standard mode only, so API content is not used to train Meta models) and in Muse Code | Meta Model API docs: Models |
| In the apps | Muse Code, Meta's coding agent, on every subscription plan: Everyday Usage ($5 a month), High Usage ($15), Power Usage ($50) and Maximum Usage ($200). Muse Wonder Full needs Power Usage or Maximum Usage | Meta: Muse Code |
| Data retention | Zero data retention for qualified Meta Model API accounts on request, though Meta says support may vary by model; calls through third-party platforms follow that platform's policy | Meta AI Developer Help Center: Does Model API support Zero Data Retention (ZDR)? |
Cached-input rates apply when part of a prompt matches a cached prefix. There is no long-context premium. Web search grounding costs $2.50 per 1,000 searches. Meta lists no batch discount, and the discounted Contributor mode is not offered for Muse Wonder on the API.
What Muse Wonder costs a month
Three typical workloads, priced at the standard API rates above. We calculated these; the provider does not publish them.
| Workload | Tokens a month | Cost a month |
|---|---|---|
| Support chatbot 100,000 replies, each 3,000 input and 400 output tokens, no caching | 300M in, 40M out | $2,000 |
| Document Q&A 5,000 questions over the same 60,000-token document set (90% cache hits), 800 output tokens each | 300M in, 4M out | $254 |
| Coding agent 200 tasks, each 1.5 million input tokens (90% cache hits) and 50,000 output tokens | 300M in, 10M out | $374 |
Reasoning models bill their thinking as output tokens, so heavy reasoning can raise the output count several times. Cache-write charges are not included.
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| Artificial Analysis Intelligence Index max effort | 52.4 | Artificial Analysis: Muse Wonder (max) |
| Coding Agent Index Muse Code harness, max effort | 63.5 | Artificial Analysis: Coding agents |
| Terminal-Bench 4.0 max effort | 52.0% | Artificial Analysis: Muse Wonder (max) |
| Humanity's Last Exam max effort | 50.1% | Artificial Analysis: Muse Wonder (max) |
| GDPval-AA v2.1 (Elo) max effort | 1770 | Artificial Analysis: Muse Wonder (max) |
| SciCode max effort | 60.3% | Artificial Analysis: Muse Wonder (max) |
Speed: about 115 output tokens a second, 36.32 seconds to the first token (max effort, Meta API; Artificial Analysis lists no default-effort variant; Artificial Analysis: Muse Wonder (max)).
Benchmark and speed data marked as such come from Artificial Analysis.
What changed
Compared with Muse Spark 1.3, input rises from $1.25 to $4 per million tokens, output from $4.25 to $20 and cached input from $0.15 to $0.20, and there is no discounted Contributor mode on the API, where Muse Spark 1.3 costs $0.10 in and $0.20 out in exchange for Meta training on your content. (Meta Model API docs: Pricing and rate limits)
Muse Wonder compared
Other models
- Gemini 4 Argon: $2 in, $10 out per million tokens
- Claude Haiku 5.5: $0.10 in, $0.50 out per million tokens
- GPT-6.1 Sol: $2 in, $10 out per million tokens
- Claude Sonnet 5.5: $2 in, $10 out per million tokens
- Claude Opus 5.5: $4 in, $20 out per million tokens
- GPT-6 Sol: $2 in, $10 out per million tokens
- GPT-6 Astra: $10 in, $50 out per million tokens
- Claude Fable 5.1: $10 in, $50 out per million tokens
- Claude Opus 5: $5 in, $25 out per million tokens
All prices side by side: AI model API price table. How we research: our method.
Sources
- Meta AI: Muse Wonder: A Frontier Model For Personal & Professional AI Agents (checked 2026-10-13)
- Meta Model API docs: Models (checked 2026-10-13)
- Meta Model API docs: Pricing and rate limits (checked 2026-10-13)
- Meta Model API docs: Reasoning (checked 2026-10-13)
- Meta: Muse Code (checked 2026-10-13)
- Meta AI Developer Help Center: Does Model API support Zero Data Retention (ZDR)? (checked 2026-10-13)
- Artificial Analysis: Muse Wonder (max) (checked 2026-10-13)
- Artificial Analysis: Coding agents (checked 2026-10-13)
- OpenRouter: Muse Wonder (checked 2026-10-13)