Model comparison
Claude Sonnet 4.6 vs Gemini 3.6 Flash: Benchmark Comparison
Claude Sonnet 4.6 has the higher score on 8 of 18 shared benchmarks; Gemini 3.6 Flash leads on 10.
The largest observed score gap is 14.96 pts on Tax Agent Bench , where Claude Sonnet 4.6 leads.
Reported ±1 standard-error ranges overlap on 6 of 18 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 Sonnet 4.6 | Gemini 3.6 Flash | Gap | Reported uncertainty |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | 5.00% ±1.65 | 3.33% ±1.17 | 1.67 pts | Reported ±1 SE ranges overlap |
| Legal Research Bench | 38.46% ±3.38 | 25.00% ±3.01 | 13.46 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 82.12% ±0.48 | 86.70% ±0.41 | 4.58 pts | Reported ±1 SE ranges do not overlap |
| EMB | 60.15% ±3.28 | 65.41% ±2.65 | 5.26 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 51.03% ±0.33 | 56.30% ±0.18 | 5.26 pts | Reported ±1 SE ranges do not overlap |
| MortgageTax | 67.73% ±0.92 | 67.61% ±0.92 | 0.12 pts | Reported ±1 SE ranges overlap |
| Tax Agent Bench | 64.85% ±3.11 | 49.90% ±3.30 | 14.96 pts | Reported ±1 SE ranges do not overlap |
| TaxEval v2 | 77.11% ±0.82 | 74.86% ±0.85 | 2.25 pts | Reported ±1 SE ranges do not overlap |
| GPQA Diamond | 85.61% ±2.41 | 93.43% ±1.33 | 7.83 pts | Reported ±1 SE ranges do not overlap |
| MMLU Pro | 87.34% ±0.43 | 89.28% ±0.30 | 1.94 pts | Reported ±1 SE ranges do not overlap |
| MMMU Pro | 83.58% ±0.89 | 88.38% ±0.77 | 4.80 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 46.58% ±3.43 | 48.69% ±3.38 | 2.10 pts | Reported ±1 SE ranges overlap |
| Code Migration | 39.89% ±4.14 | 30.93% ±4.07 | 8.96 pts | Reported ±1 SE ranges do not overlap |
| LiveCodeBench | 82.09% ±1.06 | 88.08% ±0.94 | 5.99 pts | Reported ±1 SE ranges do not overlap |
| ProgramBench | 0.50% ±0.50 | 0.00% ±0.00 | 0.50 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 77.40% ±1.87 | 79.60% ±1.80 | 2.20 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 51.48% ±4.64 | 64.00% ±4.29 | 12.53 pts | Reported ±1 SE ranges do not overlap |
| Public Benefits Bench v1.1 | 62.45% ±1.26 | 56.56% ±1.29 | 5.89 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Claude Sonnet 4.6 average | Gemini 3.6 Flash average |
|---|---|---|
| Legal | 41.86% | 38.35% |
| Finance | 64.17% | 62.81% |
| Academic | 85.51% | 90.36% |
| Education | 46.58% | 48.69% |
| Coding | 50.27% | 52.52% |
| Social Mobility | 62.45% | 56.56% |
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 Sonnet 4.6 cost | Gemini 3.6 Flash cost | Claude Sonnet 4.6 latency | Gemini 3.6 Flash latency |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | $3.04 | $1.90 | 18m43s | 16m24s |
| Legal Research Bench | $2.30 | $0.47 | 16m30s | 2m48s |
| LegalBench | N/A | N/A | 21.63s | 2.96s |
| EMB | $7.23 | $3.89 | 54m16s | 12m06s |
| Finance Agent (v2) | $2.41 | $1.40 | 11m36s | 3m45s |
| MortgageTax | N/A | N/A | 29.74s | 9.24s |
| Tax Agent Bench | $1.19 | $0.36 | 11m47s | 105.32s |
| TaxEval v2 | N/A | N/A | 2m08s | 12.34s |
| GPQA Diamond | N/A | N/A | 6m18s | 18.06s |
| MMLU Pro | N/A | N/A | 55.41s | 7.30s |
| MMMU Pro | N/A | N/A | 2m12s | 11.31s |
| SAGE | N/A | N/A | 2m43s | 21.96s |
| Code Migration | $38.51 | $8.93 | 1h40m | 1h30m |
| LiveCodeBench | N/A | N/A | 4m55s | 31.88s |
| ProgramBench | N/A | N/A | 1h12m | 47m33s |
| SWE-bench | $1.30 | $1.19 | 8m32s | 13m08s |
| Vibe Code Bench v1.1 | $5.91 | $3.04 | 26m12s | 25m23s |
| Public Benefits Bench v1.1 | $1.11 | $0.39 | 14m43s | 2m02s |
Results available only for Claude Sonnet 4.6
- SkillsBench
Results available only for Gemini 3.6 Flash
- MedCode
- MedScribe
- BioMysteryBench
- Terminal-Bench Science
- IOI
- CyberBench v1.1