Meta Muse Wonder Benchmarks and Pricing: Does It Really Beat Opus 5.5?
Meta has officially announced its latest Muse Wonder models, claiming to beat Claude Opus 5.5 and also GPT-6 Astra. Here are the detailed benchmarks of the Muse Wonder models shared by Meta and also a comparison with Claude Opus 5.5 and other models.
Muse Wonder vs Claude Opus 5.5
| Benchmark | Muse Wonder | Claude Opus 5.5 | Difference |
|---|---|---|---|
| Personal Superintelligence Index | 54.3 | 55.2 | -0.9 points |
| Harvey LAB-AA | 10.0 | 4.2 | +5.8 points |
| Vals Finance Agent V2 | 56.6 | 58.6 | -2.0 points |
| HealthBench Pro | 60.5 | 65.6 | -5.1 points |
| AutomationBench-AA | 39.1 | 41.2 | -2.1 points |
| GDP.pdf | 30.8 | 30.6 | +0.2 points |
| JobBench | 70.5 | 75.4 | -4.9 points |
| OSWorld V2.1 (partial credit) | 72.6 | 81.8 | -9.2 points |
| MemoryBench | 85.6 | 81.5 | +4.1 points |
| Terminal-Bench 4.0-AA | 52.0 | 59.6 | -7.6 points |
| DeepSWE v1.1 | 78.8 | 74.2 | +4.6 points |
| SWE-Marathon | 53.1 | 46.3 | +6.8 points |
| ProgramBench | 50.0 | 70.5 | -20.5 points |
| Artificial Analysis Index | 52 | 58 | -6 points |
Muse Wonder vs other frontier models
| Benchmark | Muse Wonder | Muse Wonder Full | Claude Opus 5.5 | GPT-6 Astra | Claude Fable 5.1 |
|---|---|---|---|---|---|
| Personal Superintelligence Index | 54.3 | 59.1 | 55.2 | 54.5 | 49.9 |
| Harvey LAB-AA | 10.0 | 13.1 | 4.2 | 8.6 | 6.4 |
| Vals Finance Agent V2 | 56.6 | 59.1 | 58.6 | 53.5 | 58.9 |
| HealthBench Pro | 60.5 | 62.3 | 65.6 | 65.3 | 62.1 |
| AutomationBench-AA | 39.1 | 45.4 | 41.2 | 41.6 | 32.1 |
| GDP.pdf | 30.8 | 33.2 | 30.6 | 34.2 | 27.6 |
| JobBench | 70.5 | 75.3 | 75.4 | 66.1 | 73.1 |
| OSWorld V2.1 (partial credit) | 72.6 | 78.4 | 81.8 | 75.8 | 80.7 |
| OSWorld V2.1 (pass or fail) | 34.3 | 39.4 | - | 34.3 | 43.1 |
| MemoryBench | 85.6 | 85.5 | 81.5 | 81.4 | 79.0 |
| Terminal-Bench 4.0-AA | 52.0 | 57.1 | 59.6 | 59.1 | 52.0 |
| DeepSWE v1.1 | 78.8 | 79.0 | 74.2 | 74.1 | 67.4 |
| SWE-Marathon | 53.1 | 57.5 (xhigh) | 46.3 | 42.5 | 45.6 |
| ProgramBench | 50.0 | 59.0 | 70.5 | 52.5 | 55.5 |
| Artificial Analysis Index | 52 | 56 | 58 | 53 | 53 |
A dash means no score has been reported for that model.
Muse Wonder vs Muse Wonder Full
| Benchmark | Muse Wonder | Muse Wonder Full | Difference |
|---|---|---|---|
| Personal Superintelligence Index | 54.3 | 59.1 | +4.8 points |
| Harvey LAB-AA | 10.0 | 13.1 | +3.1 points |
| Vals Finance Agent V2 | 56.6 | 59.1 | +2.5 points |
| HealthBench Pro | 60.5 | 62.3 | +1.8 points |
| AutomationBench-AA | 39.1 | 45.4 | +6.3 points |
| GDP.pdf | 30.8 | 33.2 | +2.4 points |
| JobBench | 70.5 | 75.3 | +4.8 points |
| OSWorld V2.1 (partial credit) | 72.6 | 78.4 | +5.8 points |
| OSWorld V2.1 (pass or fail) | 34.3 | 39.4 | +5.1 points |
| MemoryBench | 85.6 | 85.5 | -0.1 points |
| Terminal-Bench 4.0-AA | 52.0 | 57.1 | +5.1 points |
| DeepSWE v1.1 | 78.8 | 79.0 | +0.2 points |
| SWE-Marathon | 53.1 | 57.5 (xhigh) | +4.4 points |
| ProgramBench | 50.0 | 59.0 | +9.0 points |
| Artificial Analysis Index | 52 | 56 | +4 points |
API pricing
| Model | Input, per 1M tokens | Cached input | Output, per 1M tokens |
|---|---|---|---|
| Muse Wonder | $4 | $0.20 | $20 |
| Muse Wonder Full | Not on the API | Not on the API | Not on the API |
| Claude Opus 5.5 | $4 | $0.20 | $20 |
| GPT-6 Astra | $10 | - | $50 |
| Claude Fable 5.1 | $10 | - | $50 |
Muse Wonder costs the same at any context length. Web search adds $2.50 per 1,000 searches. A dash means no price is listed here.
Where you can use it
| Model | Meta Model API | Muse Code | Muse app |
|---|---|---|---|
| Muse Wonder | Available (preview) | Every plan and pay-as-you-go | Coming soon |
| Muse Wonder Full | Not available | Power Usage and Maximum Usage plans | - |
Specs
| Specification | Muse Wonder | Muse Wonder Full |
|---|---|---|
| Context window | 1M tokens (1,048,576) | 1M tokens |
| Input | Text, image, video, audio and PDF | Text, image, video and audio |
| Output | Text | Text |
| Reasoning effort | Minimal to max | - |
| Long-running agents | Yes | Yes |
| Computer use | Yes | Yes |
| Sub-agent delegation | Yes | Yes |
| Active Memory | Yes | Yes |
| Active Focus | Yes | Yes |
| API model | muse-wonder-1.0 | - |
| Release date | October 12, 2026 | October 12, 2026 |