Muse Wonder is a frontier multimodal reasoning model from Meta for personal and professional agents, long-horizon coding, and computer use. It combines Active Memory, which carries relevant context across conversations, with Active Focus, which keeps it working through blockers toward a goal.
Muse Wonder by meta is offered by 8 providers on this page. Its official list price is $4.00 per 1M input tokens and $20.00 per 1M output tokens.
Artificial Analysis rates it 52.4 on the Intelligence Index. Against its lowest blended price of $8.00 per 1M, that is roughly 7 index points per dollar, which is the value ratio the leaderboards rank on. Median output speed is 115 tokens per second, with 36.32s to the first token. Running one task of the Artificial Analysis suite costs about $4.5157, which reflects how many tokens its reasoning consumes rather than the unit price alone.
The context window is 1,048,576 tokens, with an output limit of 131,072 tokens. It supports reasoning, tool use, and structured output. Accepted input modalities are Text, Image, Video, PDF, and Audio.
Quality is independently evaluated by Artificial Analysis. Speed/latency are model-level medians.
| Provider | Tier | Input | Output | Cache read | Cache write | Context | Output limit | Status |
|---|---|---|---|---|---|---|---|---|
| LLM Gateway | Gateway | $4.00 | $20.00 | $0.20 | — | 1,048,576 | 1,048,576 | |
| MetaOfficial | First-party | $4.00 | $20.00 | $0.20 | — | 1,048,576 | 131,072 | |
| Cloudflare AI Gateway | Cloud | $4.00 | $20.00 | $0.20 | — | 1,048,576 | 131,072 | |
| Vercel AI Gateway | Cloud | $4.00 | $20.00 | $0.20 | — | 1,048,576 | 1,048,576 | |
| Requesty | Gateway | $4.00 | $20.00 | $0.20 | — | 1,048,576 | 943,718 | |
| DevPass (LLM Gateway) | Gateway | $4.00 | $20.00 | $0.20 | — | 1,048,576 | 131,072 | |
| Kilo Gateway meta/muse-wonder | Gateway | $4.00 | $20.00 | $0.20 | — | 1,048,576 | 943,718 | |
| OpenRouter meta/muse-wonder | Gateway | $4.00 | $20.00 | $0.20 | — | 1,048,576 | 943,718 |
Sorted by blended price (input×0.75 + output×0.25) asc. The official channel always shows regardless of rank. Whether a gateway's low price is actually usable can't be verified.
| Name | Conditions | Score | Metric | Source |
|---|---|---|---|---|
| Harvey Legal Agent Benchmark | variant: max effort; Artificial Analysis LAB-AA · v1.1 | 10 | all-pass rate | Source ↗ |
| Finance Agent | variant: max effort · dataset: Vals AI private test set · vv2 | 56.6 | accuracy | Source ↗ |
| HealthBench Professional | variant: max effort | 60.5 | length-normalized weighted rubric score | Source ↗ |
| AutomationBench | variant: max effort; Artificial Analysis AutomationBench-AA | 39.1 | strict pass@1 | Source ↗ |
| GDP.pdf | variant: max effort | 30.8 | score | Source ↗ |
| JobBench | variant: max effort | 70.5 | mean reward | Source ↗ |
| OSWorld | variant: max effort · v2.1 | 72.6 | mean partial score | Source ↗ |
| OSWorld | variant: max effort · v2.1 | 34.3 | binary completion rate | Source ↗ |
| Terminal-Bench | variant: max effort; Artificial Analysis Terminal-Bench 4.0-AA · v4.0 | 52 | accuracy | Source ↗ |
| DeepSWE | variant: max effort · v1.1 | 78.8 | pass@1 | Source ↗ |
| SWE-Marathon | variant: max effort | 53.1 | pass@1 | Source ↗ |
| ProgramBench | harness: mini-swe-agent · variant: max effort | 50 | share of rollouts passing at least 95% of tests | Source ↗ |
| Artificial Analysis Intelligence Index | variant: max effort · v4.3 | 52 | index score | Source ↗ |
The same benchmark scores very differently across harness / dataset, so the qualifying conditions must be shown together.