Model comparison
Gemini 3.7 Flash vs GPT-5.6 Luna: Benchmark Comparison
Gemini 3.7 Flash has the higher score on 12 of 25 shared benchmarks; GPT-5.6 Luna leads on 11.
The largest observed score gap is 29.88 pts on CyberBench v1.1 , where GPT-5.6 Luna leads.
Reported ±1 standard-error ranges overlap on 16 of 25 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 | Gemini 3.7 Flash | GPT-5.6 Luna | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 59.31% ±1.06 | 59.88% ±1.09 | 0.57 pts | Reported ±1 SE ranges overlap |
| Harvey's Legal Agent Benchmark | 8.75% ±2.15 | 1.25% ±0.83 | 7.50 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 34.62% ±3.31 | 36.54% ±3.35 | 1.92 pts | Reported ±1 SE ranges overlap |
| LegalBench | 87.26% ±0.42 | 84.03% ±0.42 | 3.22 pts | Reported ±1 SE ranges do not overlap |
| EMB | 71.33% ±2.26 | 67.12% ±2.94 | 4.21 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 59.04% ±0.27 | 55.04% ±0.31 | 4.00 pts | Reported ±1 SE ranges do not overlap |
| MortgageTax | 66.65% ±0.92 | 67.29% ±0.92 | 0.64 pts | Reported ±1 SE ranges overlap |
| Tax Agent Bench | 57.66% ±3.31 | 60.81% ±3.26 | 3.15 pts | Reported ±1 SE ranges overlap |
| TaxEval v2 | 74.73% ±0.85 | 76.17% ±0.84 | 1.43 pts | Reported ±1 SE ranges overlap |
| MedCode | 53.39% ±2.12 | 42.39% ±2.27 | 11.00 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 83.94% ±2.00 | 84.39% ±2.58 | 0.45 pts | Reported ±1 SE ranges overlap |
| ProofBench v1.1 | 58.00% ±4.96 | 60.00% ±4.92 | 2.00 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench Science | 5.71% ±2.79 | 5.71% ±2.79 | 0.00 pts | Reported ±1 SE ranges overlap |
| GPQA Diamond | 93.94% ±1.49 | 91.67% ±1.74 | 2.27 pts | Reported ±1 SE ranges overlap |
| MMLU Pro | 90.12% ±0.30 | 86.04% ±0.35 | 4.08 pts | Reported ±1 SE ranges do not overlap |
| MMMU Pro | 88.96% ±0.75 | 85.03% ±0.86 | 3.93 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 49.23% ±3.38 | 44.22% ±3.33 | 5.01 pts | Reported ±1 SE ranges overlap |
| Code Migration | 34.80% ±4.22 | 44.55% ±4.24 | 9.75 pts | Reported ±1 SE ranges do not overlap |
| IOI | 67.83% ±3.61 | 61.78% ±11.61 | 6.05 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 0.00% ±0.00 | 0.00% ±0.00 | 0.00 pts | Reported ±1 SE ranges overlap |
| SkillsBench | 65.89% ±4.55 | 60.45% ±4.75 | 5.44 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 80.80% ±1.76 | 93.00% ±1.14 | 12.20 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 6.06% ±0.88 | 4.54% ±0.88 | 1.52 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 70.39% ±4.84 | 77.06% ±3.07 | 6.66 pts | Reported ±1 SE ranges overlap |
| CyberBench v1.1 | 43.75% ±2.21 | 73.63% ±5.56 | 29.88 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Gemini 3.7 Flash average | GPT-5.6 Luna average |
|---|---|---|
| Index | 59.31% | 59.88% |
| Legal | 43.54% | 40.61% |
| Finance | 65.88% | 65.29% |
| Healthcare | 68.67% | 63.39% |
| Math | 58.00% | 60.00% |
| Science | 5.71% | 5.71% |
| Academic | 91.01% | 87.58% |
| Education | 49.23% | 44.22% |
| Coding | 46.54% | 48.77% |
| Beta | 43.75% | 73.63% |
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 | Gemini 3.7 Flash cost | GPT-5.6 Luna cost | Gemini 3.7 Flash latency | GPT-5.6 Luna latency |
|---|---|---|---|---|
| Vals Index | $4.17 | $0.78 | 43m13s | 23m16s |
| Harvey's Legal Agent Benchmark | $2.56 | $0.38 | 7m32s | 10m56s |
| Legal Research Bench | $0.76 | $0.85 | 4m49s | 39m34s |
| LegalBench | N/A | N/A | 2.23s | 5.89s |
| EMB | $5.69 | $0.37 | 10m51s | 11m00s |
| Finance Agent (v2) | $1.48 | $0.28 | 3m01s | 12m52s |
| MortgageTax | N/A | N/A | 4.11s | 22.84s |
| Tax Agent Bench | $0.48 | $0.47 | 87.24s | 42m46s |
| TaxEval v2 | N/A | N/A | 7.83s | 76.52s |
| MedCode | N/A | N/A | 9.16s | 81.28s |
| MedScribe | N/A | N/A | 16.82s | 116.89s |
| ProofBench v1.1 | $0.56 | $0.12 | 7m04s | 8m34s |
| Terminal-Bench Science | $13.22 | $0.56 | 1h08m | 1h30m |
| GPQA Diamond | N/A | N/A | 8.77s | 53.52s |
| MMLU Pro | N/A | N/A | 4.16s | 16.95s |
| MMMU Pro | N/A | N/A | 6.28s | 35.10s |
| SAGE | N/A | N/A | 9.79s | 51.88s |
| Code Migration | $21.46 | $1.88 | 2h39m | 58m35s |
| IOI | $3.40 | $0.58 | 21m09s | 1h10m |
| ProgramBench | $7.13 | $0.94 | 21m30s | 41m34s |
| SkillsBench | $1.80 | $0.21 | 3m06s | 5m20s |
| SWE-bench | $1.44 | $0.04 | 5m50s | 3m21s |
| Terminal-Bench 4.0 | $14.07 | $0.40 | 1h59m | 1h44m |
| Vibe Code Bench v1.1 | $4.83 | $0.73 | 23m43s | 25m52s |
| CyberBench v1.1 | $2.15 | $0.40 | 7m34s | 16m15s |
Results available only for Gemini 3.7 Flash
- Vals RSI Index
- LiveCodeBench
- SRE Bench
Results available only for GPT-5.6 Luna
- BioMysteryBench
- MysteryMechanism
- Vibe Code Bench 1-100
- Public Benefits Bench v1.1