Gemini 3.8 Flash vs GPT-5.6 Sol: Benchmark Comparison
Gemini 3.8 Flash has the higher score on 9 of 33 shared benchmarks; GPT-5.6 Sol leads on 24.
The largest observed score gap is 35.00 pts on ProofBench v1.1 , where GPT-5.6 Sol leads.
Reported ±1 standard-error ranges overlap on 14 of 31 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 3.8 Flash | GPT-5.6 Sol | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 54.83% ±1.03 | 58.01% ±1.02 | 3.18 pts | Reported ±1 SE ranges do not overlap |
| Vals RSI Index | 19.93% | 23.88% | 3.95 pts | Uncertainty comparison unavailable |
| Harvey's Legal Agent Benchmark | 10.00% ±2.42 | 2.50% ±0.83 | 7.50 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 38.94% ±3.39 | 48.08% ±3.47 | 9.13 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 86.99% ±0.43 | 86.97% ±0.41 | 0.03 pts | Reported ±1 SE ranges overlap |
| EMB | 72.20% ±2.42 | 72.34% ±2.37 | 0.14 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 61.44% ±0.13 | 53.76% ±0.85 | 7.68 pts | Reported ±1 SE ranges do not overlap |
| MortgageTax | 65.34% ±0.94 | 67.29% ±0.92 | 1.95 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 66.77% ±3.15 | 67.95% ±3.12 | 1.18 pts | Reported ±1 SE ranges overlap |
| TaxEval v2 | 74.45% ±0.85 | 74.78% ±0.86 | 0.33 pts | Reported ±1 SE ranges overlap |
| MedCode | 48.13% ±2.18 | 43.97% ±2.26 | 4.16 pts | Reported ±1 SE ranges overlap |
| MedScribe | 84.50% ±1.94 | 85.23% ±1.97 | 0.74 pts | Reported ±1 SE ranges overlap |
| ProofBench v1.1 | 48.00% ±5.02 | 83.00% ±3.77 | 35.00 pts | Reported ±1 SE ranges do not overlap |
| BioMysteryBench | 62.22% ±1.70 | 71.11% ±0.64 | 8.89 pts | Reported ±1 SE ranges do not overlap |
| MysteryMechanism | 36.49% ±3.24 | 33.33% ±3.17 | 3.15 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench Science | 8.57% ±3.37 | 20.00% ±4.82 | 11.43 pts | Reported ±1 SE ranges do not overlap |
| GPQA Diamond | 94.44% ±1.48 | 95.20% ±1.07 | 0.76 pts | Reported ±1 SE ranges overlap |
| MMLU Pro | 90.22% ±0.29 | 89.10% ±0.31 | 1.12 pts | Reported ±1 SE ranges do not overlap |
| MMMU Pro | 89.08% ±0.75 | 88.84% ±0.76 | 0.23 pts | Reported ±1 SE ranges overlap |
| SAGE | 35.06% ±3.36 | 52.56% ±3.42 | 17.50 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 36.55% ±4.18 | 52.92% ±4.35 | 16.37 pts | Reported ±1 SE ranges do not overlap |
| IOI | 56.94% ±2.26 | 91.17% ±4.51 | 34.22 pts | Reported ±1 SE ranges do not overlap |
| LiveCodeBench | 89.48% ±0.90 | 82.60% ±1.09 | 6.88 pts | Reported ±1 SE ranges do not overlap |
| ProgramBench | 1.00% ±0.70 | 1.50% ±0.86 | 0.50 pts | Reported ±1 SE ranges overlap |
| SkillsBench | 57.98% ±4.42 | 54.10% ±4.73 | 3.88 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 80.00% ±1.79 | 96.20% ±0.86 | 16.20 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 19.19% ±2.52 | 37.88% ±0.00 | 18.69 pts | Reported ±1 SE ranges do not overlap |
| Vibe Code Bench 1-100 | 18.77% ±3.31 | 20.01% ±3.63 | 1.24 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 78.65% ±3.88 | 80.50% ±3.72 | 1.84 pts | Reported ±1 SE ranges overlap |
| CyberBench v1.1 | 43.75% ±2.21 | 76.31% ±5.18 | 32.56 pts | Reported ±1 SE ranges do not overlap |
| Public Benefits Bench v1.1 | 65.29% ±1.24 | 66.51% ±1.23 | 1.22 pts | Reported ±1 SE ranges overlap |
| CUA-bench | 4.17% | 8.33% | 4.17 pts | Uncertainty comparison unavailable |
| Time Horizon Index: KSP | 18.83% ±0.00 | 23.83% ±0.00 | 5.00 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Gemini 3.8 Flash average | GPT-5.6 Sol average |
|---|---|---|
| Legal | 45.31% | 45.85% |
| Finance | 68.04% | 67.22% |
| Healthcare | 66.32% | 64.60% |
| Math | 48.00% | 83.00% |
| Science | 35.76% | 41.48% |
| Academic | 91.25% | 91.05% |
| Education | 35.06% | 52.56% |
| Coding | 48.73% | 57.43% |
| Cyber | 43.75% | 76.31% |
| Social Mobility | 65.29% | 66.51% |
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 3.8 Flash cost | GPT-5.6 Sol cost | Gemini 3.8 Flash latency | GPT-5.6 Sol latency |
|---|---|---|---|---|
| Vals Index | $5.73 | $14.24 | 51m57s | 37m23s |
| Vals RSI Index | $332.04 | $347.10 | 90h00m | 90h00m |
| Harvey's Legal Agent Benchmark | $3.66 | $10.37 | 29m21s | 22m50s |
| Legal Research Bench | $1.63 | $21.61 | 5m24s | 1h17m |
| LegalBench | N/A | N/A | 3.32s | 6.20s |
| EMB | $8.24 | $6.01 | 12m49s | 9m50s |
| Finance Agent (v2) | $2.00 | $1.25 | 3m21s | 19m25s |
| MortgageTax | N/A | N/A | 14.79s | 12.62s |
| Tax Agent Bench | $0.86 | $7.51 | 2m58s | 44m41s |
| TaxEval v2 | N/A | N/A | 8.76s | 62.03s |
| MedCode | N/A | N/A | 42.89s | 96.59s |
| MedScribe | N/A | N/A | 21.14s | 94.40s |
| ProofBench v1.1 | $0.60 | $1.44 | 6m09s | 8m40s |
| BioMysteryBench | $1.55 | $1.93 | 5m27s | 18m18s |
| MysteryMechanism | $1.73 | $0.96 | 5m22s | 8m43s |
| Terminal-Bench Science | $5.64 | $7.32 | 55m11s | 1h44m |
| GPQA Diamond | N/A | N/A | 20.18s | 58.30s |
| MMLU Pro | N/A | N/A | 7.82s | 15.65s |
| MMMU Pro | N/A | N/A | 12.03s | 26.20s |
| SAGE | N/A | N/A | 28.64s | 75.39s |
| Code Migration | $18.49 | $24.54 | 2h37m | 59m12s |
| IOI | $3.98 | $7.98 | 14m20s | 1h02m |
| LiveCodeBench | N/A | N/A | 24.07s | 56.51s |
| ProgramBench | $10.93 | $15.29 | 46m39s | 29m24s |
| SkillsBench | $2.34 | $4.14 | 3m55s | 13m07s |
| SWE-bench | $2.19 | $1.15 | 11m23s | 3m02s |
| Terminal-Bench 4.0 | $8.77 | $7.98 | 1h48m | 33m46s |
| Vibe Code Bench 1-100 | $18.91 | $47.43 | 1h06m | 2h37m |
| Vibe Code Bench v1.1 | $6.87 | $33.40 | 8m39s | 33m53s |
| CyberBench v1.1 | $1.31 | $3.05 | 9m49s | 10m47s |
| Public Benefits Bench v1.1 | $1.01 | $8.85 | 3m21s | 1h01m |
| CUA-bench | $280.03 | $206.99 | 36.55s | 18.80s |
| Time Horizon Index: KSP | $543.14 | $2027.20 | N/A | N/A |
Results available only for Gemini 3.8 Flash
None.
Results available only for GPT-5.6 Sol
- SRE Bench