Claude Fable 5 vs GPT-5.6 Luna: Benchmark Comparison
Claude Fable 5 has the higher score on 23 of 23 shared benchmarks; GPT-5.6 Luna leads on 0.
The largest observed score gap is 35.00 pts on ProofBench v1.1 , where Claude Fable 5 leads.
Reported ±1 standard-error ranges overlap on 5 of 23 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-5.6 Luna | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 61.39% ±1.00 | 51.69% ±1.06 | 9.70 pts | Reported ±1 SE ranges do not overlap |
| Harvey's Legal Agent Benchmark | 11.25% ±2.15 | 1.25% ±0.83 | 10.00 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 49.52% ±3.48 | 36.54% ±3.35 | 12.98 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 88.56% ±0.33 | 84.03% ±0.42 | 4.53 pts | Reported ±1 SE ranges do not overlap |
| EMB | 73.67% ±2.47 | 67.12% ±2.94 | 6.55 pts | Reported ±1 SE ranges do not overlap |
| Finance Agent (v2) | 56.31% ±0.84 | 55.04% ±0.31 | 1.27 pts | Reported ±1 SE ranges do not overlap |
| MortgageTax | 68.92% ±0.91 | 67.29% ±0.92 | 1.63 pts | Reported ±1 SE ranges overlap |
| Tax Agent Bench | 69.82% ±3.12 | 60.81% ±3.26 | 9.01 pts | Reported ±1 SE ranges do not overlap |
| TaxEval v2 | 76.94% ±0.82 | 76.17% ±0.84 | 0.78 pts | Reported ±1 SE ranges overlap |
| MedCode | 56.07% ±2.20 | 42.39% ±2.27 | 13.68 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 88.52% ±1.95 | 84.39% ±2.58 | 4.13 pts | Reported ±1 SE ranges overlap |
| ProofBench v1.1 | 95.00% ±2.19 | 60.00% ±4.92 | 35.00 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench Science | 15.71% ±4.38 | 0.00% ±0.00 | 15.71 pts | Reported ±1 SE ranges do not overlap |
| GPQA Diamond | 93.18% ±1.94 | 91.67% ±1.74 | 1.52 pts | Reported ±1 SE ranges overlap |
| MMLU Pro | 91.50% ±0.28 | 86.04% ±0.35 | 5.47 pts | Reported ±1 SE ranges do not overlap |
| MMMU Pro | 89.31% ±0.74 | 85.03% ±0.86 | 4.28 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 51.89% ±3.40 | 44.22% ±3.33 | 7.67 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 55.06% ±4.61 | 44.55% ±4.24 | 10.52 pts | Reported ±1 SE ranges do not overlap |
| ProgramBench | 2.00% ±0.99 | 0.00% ±0.00 | 2.00 pts | Reported ±1 SE ranges do not overlap |
| SWE-bench | 95.00% ±0.98 | 93.00% ±1.14 | 2.00 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench 4.0 | 41.41% ±1.34 | 11.62% ±1.01 | 29.80 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 90.35% ±2.10 | 77.06% ±3.07 | 13.30 pts | Reported ±1 SE ranges do not overlap |
| Public Benefits Bench v1.1 | 70.43% ±1.19 | 61.16% ±1.27 | 9.27 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Claude Fable 5 average | GPT-5.6 Luna average |
|---|---|---|
| Legal | 49.78% | 40.61% |
| Finance | 69.13% | 65.29% |
| Healthcare | 72.30% | 63.39% |
| Math | 95.00% | 60.00% |
| Science | 15.71% | 0.00% |
| Academic | 91.33% | 87.58% |
| Education | 51.89% | 44.22% |
| Coding | 56.77% | 45.24% |
| Social Mobility | 70.43% | 61.16% |
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-5.6 Luna cost | Claude Fable 5 latency | GPT-5.6 Luna latency |
|---|---|---|---|---|
| Vals Index | $29.59 | $0.82 | 42m24s | 29m27s |
| Harvey's Legal Agent Benchmark | $19.23 | $0.38 | 26m53s | 10m56s |
| Legal Research Bench | $9.79 | $0.85 | 22m18s | 39m34s |
| LegalBench | N/A | N/A | 8.96s | 5.89s |
| EMB | $12.35 | $0.37 | 26m34s | 11m00s |
| Finance Agent (v2) | $8.06 | $0.28 | 10m11s | 12m52s |
| MortgageTax | N/A | N/A | 16.10s | 22.84s |
| Tax Agent Bench | $6.70 | $0.47 | 14m27s | 42m46s |
| TaxEval v2 | N/A | N/A | 56.87s | 76.52s |
| MedCode | N/A | N/A | 91.44s | 81.28s |
| MedScribe | N/A | N/A | 119.47s | 116.89s |
| ProofBench v1.1 | $5.45 | $0.12 | 13m43s | 8m34s |
| Terminal-Bench Science | $51.99 | $0.58 | 1h59m | 2h03m |
| GPQA Diamond | N/A | N/A | 99.90s | 53.52s |
| MMLU Pro | N/A | N/A | 25.00s | 16.95s |
| MMMU Pro | N/A | N/A | 61.44s | 35.10s |
| SAGE | N/A | N/A | 116.92s | 51.88s |
| Code Migration | $112.10 | $1.88 | 1h49m | 58m35s |
| ProgramBench | $75.68 | $0.94 | 2h37m | 41m34s |
| SWE-bench | $2.05 | $0.04 | 5m56s | 3m21s |
| Terminal-Bench 4.0 | $30.34 | $0.73 | 1h08m | 35m55s |
| Vibe Code Bench v1.1 | $41.71 | $0.73 | 1h02m | 25m52s |
| Public Benefits Bench v1.1 | $4.59 | $0.27 | 21m17s | 38m40s |
Results available only for Claude Fable 5
- Vals RSI Index
- LiveCodeBench
Results available only for GPT-5.6 Luna
- BioMysteryBench
- MysteryMechanism
- IOI
- SkillsBench
- Vibe Code Bench 1-100
- CyberBench v1.1