Claude Opus 4.8 vs GPT-5.6 Luna: Benchmark Comparison
Claude Opus 4.8 has the higher score on 18 of 23 shared benchmarks; GPT-5.6 Luna leads on 5.
The largest observed score gap is 11.62 pts on Terminal-Bench 4.0 , where Claude Opus 4.8 leads.
Reported ±1 standard-error ranges overlap on 10 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 Opus 4.8 | GPT-5.6 Luna | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 55.10% ±1.01 | 51.69% ±1.06 | 3.41 pts | Reported ±1 SE ranges do not overlap |
| Harvey's Legal Agent Benchmark | 9.58% ±2.15 | 1.25% ±0.83 | 8.33 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 43.75% ±3.45 | 36.54% ±3.35 | 7.21 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 83.57% ±0.47 | 84.03% ±0.42 | 0.47 pts | Reported ±1 SE ranges overlap |
| EMB | 69.37% ±2.61 | 67.12% ±2.94 | 2.26 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 53.92% ±0.16 | 55.04% ±0.31 | 1.13 pts | Reported ±1 SE ranges do not overlap |
| MortgageTax | 69.91% ±0.89 | 67.29% ±0.92 | 2.62 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 64.91% ±3.17 | 60.81% ±3.26 | 4.09 pts | Reported ±1 SE ranges overlap |
| TaxEval v2 | 75.63% ±0.84 | 76.17% ±0.84 | 0.53 pts | Reported ±1 SE ranges overlap |
| MedCode | 53.22% ±2.17 | 42.39% ±2.27 | 10.83 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 85.75% ±1.93 | 84.39% ±2.58 | 1.36 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench Science | 4.29% ±2.44 | 0.00% ±0.00 | 4.29 pts | Reported ±1 SE ranges do not overlap |
| GPQA Diamond | 92.42% ±1.98 | 91.67% ±1.74 | 0.76 pts | Reported ±1 SE ranges overlap |
| MMLU Pro | 89.58% ±0.30 | 86.04% ±0.35 | 3.55 pts | Reported ±1 SE ranges do not overlap |
| MMMU Pro | 86.59% ±0.82 | 85.03% ±0.86 | 1.56 pts | Reported ±1 SE ranges overlap |
| SAGE | 54.79% ±3.35 | 44.22% ±3.33 | 10.57 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 47.25% ±4.18 | 44.55% ±4.24 | 2.70 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 1.00% ±0.70 | 0.00% ±0.00 | 1.00 pts | Reported ±1 SE ranges do not overlap |
| SkillsBench | 59.23% ±4.50 | 60.45% ±4.75 | 1.22 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 88.60% ±1.42 | 93.00% ±1.14 | 4.40 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 23.23% ±1.34 | 11.62% ±1.01 | 11.62 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench v1.1 | 82.72% ±3.08 | 77.06% ±3.07 | 5.67 pts | Reported ±1 SE ranges overlap |
| Public Benefits Bench v1.1 | 68.13% ±1.21 | 61.16% ±1.27 | 6.97 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Claude Opus 4.8 average | GPT-5.6 Luna average |
|---|---|---|
| Legal | 45.63% | 40.61% |
| Finance | 66.75% | 65.29% |
| Healthcare | 69.49% | 63.39% |
| Science | 4.29% | 0.00% |
| Academic | 89.53% | 87.58% |
| Education | 54.79% | 44.22% |
| Coding | 50.34% | 47.78% |
| Social Mobility | 68.13% | 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 Opus 4.8 cost | GPT-5.6 Luna cost | Claude Opus 4.8 latency | GPT-5.6 Luna latency |
|---|---|---|---|---|
| Vals Index | $13.14 | $0.82 | 39m48s | 29m27s |
| Harvey's Legal Agent Benchmark | $10.22 | $0.38 | 24m08s | 10m56s |
| Legal Research Bench | $2.82 | $0.85 | 11m54s | 39m34s |
| LegalBench | N/A | N/A | 3.11s | 5.89s |
| EMB | $12.06 | $0.37 | 43m52s | 11m00s |
| Finance Agent (v2) | $4.22 | $0.28 | 9m07s | 12m52s |
| MortgageTax | N/A | N/A | 20.81s | 22.84s |
| Tax Agent Bench | $1.69 | $0.47 | 7m59s | 42m46s |
| TaxEval v2 | N/A | N/A | 70.33s | 76.52s |
| MedCode | N/A | N/A | 105.92s | 81.28s |
| MedScribe | N/A | N/A | 82.36s | 116.89s |
| Terminal-Bench Science | $23.14 | $0.58 | 1h55m | 2h03m |
| GPQA Diamond | N/A | N/A | 2m14s | 53.52s |
| MMLU Pro | N/A | N/A | 23.44s | 16.95s |
| MMMU Pro | N/A | N/A | 49.04s | 35.10s |
| SAGE | N/A | N/A | 3m06s | 51.88s |
| Code Migration | $30.51 | $1.88 | 1h15m | 58m35s |
| ProgramBench | $31.27 | $0.94 | 1h32m | 41m34s |
| SkillsBench | $4.77 | $0.21 | 15m30s | 5m20s |
| SWE-bench | $1.92 | $0.04 | 9m27s | 3m21s |
| Terminal-Bench 4.0 | $17.14 | $0.73 | 1h11m | 35m55s |
| Vibe Code Bench v1.1 | $26.88 | $0.73 | 1h16m | 25m52s |
| Public Benefits Bench v1.1 | $1.89 | $0.27 | 12m48s | 38m40s |
Results available only for Claude Opus 4.8
- Vals RSI Index
- LiveCodeBench
- Time Horizon Index: KSP
Results available only for GPT-5.6 Luna
- ProofBench v1.1
- BioMysteryBench
- MysteryMechanism
- IOI
- Vibe Code Bench 1-100
- CyberBench v1.1