Claude Fable 5 vs GPT-6 Sol: Benchmark Comparison
Claude Fable 5 has the higher score on 12 of 17 shared benchmarks; GPT-6 Sol leads on 4.
The largest observed score gap is 20.67 pts on Legal Research Bench , where Claude Fable 5 leads.
Reported ±1 standard-error ranges overlap on 5 of 16 shared benchmarks with comparable uncertainty data. This is not a pairwise statistical significance test.
Shared benchmark results
Scores and reported standard errors come from the latest visible Vals benchmark version available for both models.
| Benchmark | Claude Fable 5 | GPT-6 Sol | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 61.39% ±1.00 | 57.54% ±1.01 | 3.85 pts | Reported ±1 SE ranges do not overlap |
| Vals RSI Index | 24.32% | 28.11% | 3.79 pts | Uncertainty comparison unavailable |
| Harvey's Legal Agent Benchmark | 11.25% ±2.15 | 1.67% ±0.82 | 9.58 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 49.52% ±3.48 | 28.85% ±3.15 | 20.67 pts | Reported ±1 SE ranges do not overlap |
| EMB | 73.67% ±2.47 | 71.53% ±2.35 | 2.14 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 56.31% ±0.84 | 49.05% ±0.58 | 7.26 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 69.82% ±3.12 | 53.05% ±3.29 | 16.78 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 56.07% ±2.20 | 47.07% ±2.12 | 9.00 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 88.52% ±1.95 | 82.03% ±1.94 | 6.49 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 95.00% ±2.19 | 83.00% ±3.77 | 12.00 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench Science | 15.71% ±4.38 | 30.00% ±5.52 | 14.29 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 51.89% ±3.40 | 44.79% ±3.36 | 7.10 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 55.06% ±4.61 | 57.20% ±4.21 | 2.13 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 2.00% ±0.99 | 2.00% ±0.99 | 0.00 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench 4.0 | 41.41% ±1.34 | 44.44% ±3.54 | 3.03 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 90.35% ±2.10 | 87.82% ±2.53 | 2.53 pts | Reported ±1 SE ranges overlap |
| Public Benefits Bench v1.1 | 70.43% ±1.19 | 56.63% ±1.29 | 13.80 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Claude Fable 5 average | GPT-6 Sol average |
|---|---|---|
| Legal | 30.38% | 15.26% |
| Finance | 66.60% | 57.88% |
| Healthcare | 72.30% | 64.55% |
| Math | 95.00% | 83.00% |
| Science | 15.71% | 30.00% |
| Education | 51.89% | 44.79% |
| Coding | 47.21% | 47.87% |
| Social Mobility | 70.43% | 56.63% |
Cost and latency
Cost per test appears only where the benchmark reports it for both models. Latency is the measured completion time for that benchmark.
| Benchmark | Claude Fable 5 cost | GPT-6 Sol cost | Claude Fable 5 latency | GPT-6 Sol latency |
|---|---|---|---|---|
| Vals Index | $29.59 | $7.58 | 42m24s | 29m11s |
| Vals RSI Index | $491.22 | $665.11 | 90h00m | 90h00m |
| Harvey's Legal Agent Benchmark | $19.23 | $3.36 | 26m53s | 13m40s |
| Legal Research Bench | $9.79 | $4.90 | 22m18s | 24m09s |
| EMB | $12.35 | $1.28 | 26m34s | 10m22s |
| Finance Agent (v2) | $8.06 | $2.12 | 10m11s | 9m00s |
| Tax Agent Bench | $6.70 | $2.28 | 14m27s | 17m32s |
| MedCode | N/A | N/A | 91.44s | 62.71s |
| MedScribe | N/A | N/A | 119.47s | 81.14s |
| ProofBench v1.1 | $5.45 | $0.48 | 13m43s | 5m05s |
| Terminal-Bench Science | $51.99 | $5.82 | 1h59m | 1h22m |
| SAGE | N/A | N/A | 116.92s | 31.42s |
| Code Migration | $112.10 | $15.68 | 1h49m | 1h26m |
| ProgramBench | $75.68 | $11.67 | 2h37m | 46m11s |
| Terminal-Bench 4.0 | $30.34 | $5.79 | 1h08m | 34m56s |
| Vibe Code Bench v1.1 | $41.71 | $26.36 | 1h02m | 38m56s |
| Public Benefits Bench v1.1 | $4.59 | $2.49 | 21m17s | 31m02s |
Results available only for Claude Fable 5
- LegalBench
- MortgageTax
- TaxEval v2
- GPQA Diamond
- MMLU Pro
- MMMU Pro
- LiveCodeBench
- SWE-bench
Results available only for GPT-6 Sol
- BioMysteryBench
- MysteryMechanism
- IOI
- CyberBench v1.1