Model comparison
Claude Sonnet 4.6 vs Gemini 3.5 Flash: Benchmark Comparison
Claude Sonnet 4.6 has the higher score on 8 of 19 shared benchmarks; Gemini 3.5 Flash leads on 11.
The largest observed score gap is 13.15 pts on Code Migration , where Claude Sonnet 4.6 leads.
Reported ±1 standard-error ranges overlap on 7 of 19 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 3.5 Flash | Gap | Reported uncertainty |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | 5.00% ±1.65 | 2.50% ±0.83 | 2.50 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 38.46% ±3.38 | 30.77% ±3.21 | 7.69 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 82.12% ±0.48 | 83.60% ±0.85 | 1.48 pts | Reported ±1 SE ranges do not overlap |
| EMB | 60.15% ±3.28 | 63.55% ±2.75 | 3.39 pts | Reported ±1 SE ranges overlap |
| Finance Agent (v2) | 51.03% ±0.33 | 57.86% ±0.23 | 6.83 pts | Reported ±1 SE ranges do not overlap |
| MortgageTax | 67.73% ±0.92 | 68.12% ±0.91 | 0.40 pts | Reported ±1 SE ranges overlap |
| Tax Agent Bench | 64.85% ±3.11 | 53.53% ±3.30 | 11.33 pts | Reported ±1 SE ranges do not overlap |
| TaxEval v2 | 77.11% ±0.82 | 74.37% ±0.85 | 2.74 pts | Reported ±1 SE ranges do not overlap |
| GPQA Diamond | 85.61% ±2.41 | 92.68% ±1.46 | 7.07 pts | Reported ±1 SE ranges do not overlap |
| MMLU Pro | 87.34% ±0.43 | 89.52% ±0.31 | 2.17 pts | Reported ±1 SE ranges do not overlap |
| MMMU Pro | 83.58% ±0.89 | 88.27% ±0.77 | 4.68 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 46.58% ±3.43 | 49.88% ±3.44 | 3.30 pts | Reported ±1 SE ranges overlap |
| Code Migration | 39.89% ±4.14 | 26.75% ±4.10 | 13.15 pts | Reported ±1 SE ranges do not overlap |
| LiveCodeBench | 82.09% ±1.06 | 87.60% ±0.95 | 5.51 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 |
| SkillsBench | 49.05% ±4.64 | 52.74% ±4.47 | 3.69 pts | Reported ±1 SE ranges overlap |
| SWE-bench | 77.40% ±1.87 | 78.80% ±1.83 | 1.40 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 51.48% ±4.64 | 48.68% ±4.73 | 2.79 pts | Reported ±1 SE ranges overlap |
| Public Benefits Bench v1.1 | 62.45% ±1.26 | 59.47% ±1.28 | 2.98 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Claude Sonnet 4.6 average | Gemini 3.5 Flash average |
|---|---|---|
| Legal | 41.86% | 38.96% |
| Finance | 64.17% | 63.48% |
| Academic | 85.51% | 90.15% |
| Education | 46.58% | 49.88% |
| Coding | 50.07% | 49.09% |
| Social Mobility | 62.45% | 59.47% |
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 3.5 Flash cost | Claude Sonnet 4.6 latency | Gemini 3.5 Flash latency |
|---|---|---|---|---|
| Harvey's Legal Agent Benchmark | $3.04 | $2.12 | 18m43s | 7m58s |
| Legal Research Bench | $2.30 | $1.31 | 16m30s | 4m38s |
| LegalBench | N/A | N/A | 21.63s | 34.58s |
| EMB | $7.23 | $5.17 | 54m16s | 18m44s |
| Finance Agent (v2) | $2.41 | $2.51 | 11m36s | 5m22s |
| MortgageTax | N/A | N/A | 29.74s | 11.43s |
| Tax Agent Bench | $1.19 | $0.86 | 11m47s | 3m53s |
| TaxEval v2 | N/A | N/A | 2m08s | 19.69s |
| GPQA Diamond | N/A | N/A | 6m18s | 27.66s |
| MMLU Pro | N/A | N/A | 55.41s | 26.26s |
| MMMU Pro | N/A | N/A | 2m12s | 12.23s |
| SAGE | N/A | N/A | 2m43s | 32.59s |
| Code Migration | $38.51 | $5.73 | 1h40m | 19m28s |
| LiveCodeBench | N/A | N/A | 4m55s | 64.37s |
| ProgramBench | N/A | N/A | 1h12m | 18m25s |
| SkillsBench | $1.38 | $1.25 | 14m20s | 7m15s |
| SWE-bench | $1.30 | $0.95 | 8m32s | 4m14s |
| Vibe Code Bench v1.1 | $5.91 | $2.54 | 26m12s | 14m53s |
| Public Benefits Bench v1.1 | $1.11 | $1.11 | 14m43s | 5m39s |
Results available only for Claude Sonnet 4.6
None.
Results available only for Gemini 3.5 Flash
- Vals Index
- MedCode
- MedScribe
- ProofBench v1.1
- Terminal-Bench Science
- Terminal-Bench 4.0