Muse Wonder vs GPT-6.1 Sol: Did Meta Beat OpenAI?
Claude Opus 5.5 is a huge leap, and even GPT-6.1 Sol has been left behind. However, Meta's latest Muse Wonder flagship models claim to beat Opus 5.5, and here is how Muse Wonder compares with the latest GPT-6.1 Sol.
Benchmarks
| Benchmark | Muse Wonder | GPT-6.1 Sol | Difference |
|---|---|---|---|
| DeepSWE v1.1 | 78.8%Max effort | 75.2%High effort | Wonder +3.6 points |
| GDP.pdf | 30.8%Max effort | 32.0%High effort | Sol +1.2 points |
| AutomationBench | 39.1%AutomationBench-AA, max effort | 36.1%AutomationBench 1.0.6, max effort | Wonder +3.0 points |
Artificial Analysis benchmarks
| Benchmark | Muse Wonder (Max) | GPT-6.1 Sol (Max) | Difference |
|---|---|---|---|
| Intelligence Index v4.3.2 | 52 | 52 | Tie |
| AA-Briefcase v1.1 | 1659 Elo | 1557 Elo | Wonder +102 Elo |
| GDPval-AA v2.1 | 1770 Elo | 1592 Elo | Wonder +178 Elo |
| AutomationBench-AA | 68% | 65% | Wonder +3 points |
| Terminal-Bench 4.0 | 52% | 56% | Sol +4 points |
| SciCode | 60% | 54% | Wonder +6 points |
| Humanity's Last Exam | 50% | 53% | Sol +3 points |
| GDP.pdf | 31% | 31% | Tie |
| CritPt | 28% | 32% | Sol +4 points |
| AA-Omniscience | 25 | 42 | Sol +17 |
| AA-LCR v1.1 | 81% | 83% | Sol +2 points |
Cost and efficiency
| Metric | Muse Wonder (Max) | GPT-6.1 Sol (Max) | Difference |
|---|---|---|---|
| Intelligence Index | 52 | 52 | Tie |
| Cost per Intelligence Index task | $4.52 | $0.72 | Sol about 84% cheaper |
| Cost to run the Intelligence Index | $5,733 | $1,082 | Sol about 81% cheaper |
| Output tokens per task | 54K | 38K | Sol uses about 30% fewer |
| Reasoning tokens per task | 29K | 25K | Sol uses about 12% fewer |
| Output tokens across the Index | 130M | 67M | Sol uses about 48% fewer |
| Output speed | 85 tokens/s | 65 tokens/s | Wonder about 32% faster |
| Time to first token | 37.4 s | 339.4 s | Wonder starts much sooner |
Pricing and specs
| Specification | Muse Wonder | GPT-6.1 Sol | Difference |
|---|---|---|---|
| Input, per 1M tokens | $4 | $2 | Sol 50% cheaper |
| Cached input, per 1M tokens | $0.20 | $0.10 | Sol 50% cheaper |
| Output, per 1M tokens | $20 | $10 | Sol 50% cheaper |
| Long-context surcharge | None | Yes, above 272K input tokens | Wonder |
| Context window | 1,048,576 tokens | 1.05M tokens | Nearly the same |
| Max output | Not stated | 128K tokens | - |
| Reasoning effort | Minimal to max | Low to max | Both |
| Text input | Yes | Yes | Same |
| Image input | Yes | Yes | Same |
| Audio input | Yes | No | Wonder |
| Video input | Yes | No | Wonder |
| API | Meta Model API (preview) | OpenAI API | Both |