Claude Opus 5.5 vs GPT-5.6 Luna: Benchmark Comparison
Claude Opus 5.5 has the higher score on 20 of 21 shared benchmarks; GPT-5.6 Luna leads on 1.
The largest observed score gap is 53.54 pts on Terminal-Bench 4.0 , where Claude Opus 5.5 leads.
Reported ±1 standard-error ranges overlap on 2 of 21 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 5.5 | GPT-5.6 Luna | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 66.97% ±0.89 | 51.69% ±1.06 | 15.28 pts | Reported ±1 SE ranges do not overlap |
| Harvey's Legal Agent Benchmark | 3.75% ±1.47 | 1.25% ±0.83 | 2.50 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 50.48% ±3.48 | 36.54% ±3.35 | 13.94 pts | Reported ±1 SE ranges do not overlap |
| EMB | 75.94% ±2.38 | 67.12% ±2.94 | 8.82 pts | Reported ±1 SE ranges do not overlap |
| Finance Agent (v2) | 58.59% ±0.17 | 55.04% ±0.31 | 3.54 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 70.50% ±3.15 | 60.81% ±3.26 | 9.68 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 49.80% ±2.27 | 42.39% ±2.27 | 7.41 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 91.43% ±1.93 | 84.39% ±2.58 | 7.04 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 100.00% ±0.00 | 60.00% ±4.92 | 40.00 pts | Reported ±1 SE ranges do not overlap |
| BioMysteryBench | 79.26% ±0.98 | 61.48% ±0.74 | 17.78 pts | Reported ±1 SE ranges do not overlap |
| MysteryMechanism | 49.55% ±3.36 | 14.41% ±2.36 | 35.14 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench Science | 47.14% ±6.01 | 0.00% ±0.00 | 47.14 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 45.83% ±3.36 | 44.22% ±3.33 | 1.61 pts | Reported ±1 SE ranges overlap |
| Code Migration | 66.65% ±4.33 | 44.55% ±4.24 | 22.10 pts | Reported ±1 SE ranges do not overlap |
| IOI | 95.06% ±4.94 | 61.78% ±11.61 | 33.28 pts | Reported ±1 SE ranges do not overlap |
| ProgramBench | 18.50% ±2.75 | 0.00% ±0.00 | 18.50 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 65.15% ±0.00 | 11.62% ±1.01 | 53.54 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench 1-100 | 30.36% ±4.83 | 22.59% ±4.00 | 7.77 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 90.29% ±1.53 | 77.06% ±3.07 | 13.24 pts | Reported ±1 SE ranges do not overlap |
| CyberBench v1.1 | 55.36% ±5.13 | 73.63% ±5.56 | 18.27 pts | Reported ±1 SE ranges do not overlap |
| Public Benefits Bench v1.1 | 70.64% ±1.19 | 61.16% ±1.27 | 9.47 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Claude Opus 5.5 average | GPT-5.6 Luna average |
|---|---|---|
| Legal | 27.12% | 18.89% |
| Finance | 68.34% | 60.99% |
| Healthcare | 70.61% | 63.39% |
| Math | 100.00% | 60.00% |
| Science | 58.65% | 25.30% |
| Education | 45.83% | 44.22% |
| Coding | 61.00% | 36.27% |
| Cyber | 55.36% | 73.63% |
| Social Mobility | 70.64% | 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 5.5 cost | GPT-5.6 Luna cost | Claude Opus 5.5 latency | GPT-5.6 Luna latency |
|---|---|---|---|---|
| Vals Index | $32.14 | $0.82 | 1h19m | 29m27s |
| Harvey's Legal Agent Benchmark | $21.38 | $0.38 | 49m00s | 10m56s |
| Legal Research Bench | $25.16 | $0.85 | 1h52m | 39m34s |
| EMB | $11.02 | $0.37 | 35m06s | 11m00s |
| Finance Agent (v2) | $9.22 | $0.28 | 33m08s | 12m52s |
| Tax Agent Bench | $15.09 | $0.47 | 1h07m | 42m46s |
| MedCode | N/A | N/A | 4m07s | 81.28s |
| MedScribe | N/A | N/A | 7m22s | 116.89s |
| ProofBench v1.1 | $0.96 | $0.12 | 5m51s | 8m34s |
| BioMysteryBench | $3.64 | $0.09 | 12m19s | 18m31s |
| MysteryMechanism | $4.62 | $0.11 | 19m38s | 8m13s |
| Terminal-Bench Science | $19.12 | $0.58 | 2h28m | 2h03m |
| SAGE | N/A | N/A | 114.20s | 51.88s |
| Code Migration | $112.97 | $1.88 | 3h19m | 58m35s |
| IOI | $5.26 | $0.58 | 17m15s | 1h10m |
| ProgramBench | $68.05 | $0.94 | 2h24m | 41m34s |
| Terminal-Bench 4.0 | $13.20 | $0.73 | 1h04m | 35m55s |
| Vibe Code Bench 1-100 | $44.12 | $4.91 | 2h28m | 2h03m |
| Vibe Code Bench v1.1 | $57.92 | $0.73 | 1h36m | 25m52s |
| CyberBench v1.1 | $3.48 | $0.40 | 18m11s | 16m15s |
| Public Benefits Bench v1.1 | $5.95 | $0.27 | 1h20m | 38m40s |
Results available only for Claude Opus 5.5
- Vals RSI Index
- SRE Bench
- CUA-bench
- Time Horizon Index: KSP
Results available only for GPT-5.6 Luna
- LegalBench
- MortgageTax
- TaxEval v2
- GPQA Diamond
- MMLU Pro
- MMMU Pro
- SkillsBench
- SWE-bench