Gemini 3.5 Flash vs Grok 4.5: Benchmark Comparison
Gemini 3.5 Flash has the higher score on 10 of 23 shared benchmarks; Grok 4.5 leads on 11.
The largest observed score gap is 26.47 pts on MMMU Pro , where Gemini 3.5 Flash leads.
Reported ±1 standard-error ranges overlap on 7 of 23 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 3.5 Flash | Grok 4.5 | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 44.79% ±1.10 | 44.72% ±1.17 | 0.06 pts | Reported ±1 SE ranges overlap |
| Harvey's Legal Agent Benchmark | 2.50% ±0.83 | 12.92% ±2.53 | 10.42 pts | Reported ±1 SE ranges do not overlap |
| Legal Research Bench | 30.77% ±3.21 | 37.98% ±3.37 | 7.21 pts | Reported ±1 SE ranges do not overlap |
| LegalBench | 83.60% ±0.85 | 85.97% ±0.41 | 2.36 pts | Reported ±1 SE ranges do not overlap |
| EMB | 63.55% ±2.75 | 52.91% ±3.11 | 10.64 pts | Reported ±1 SE ranges do not overlap |
| Finance Agent (v2) | 57.86% ±0.23 | 48.35% ±0.46 | 9.51 pts | Reported ±1 SE ranges do not overlap |
| MortgageTax | 68.12% ±0.91 | 61.80% ±0.97 | 6.32 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 53.53% ±3.30 | 64.32% ±3.16 | 10.80 pts | Reported ±1 SE ranges do not overlap |
| TaxEval v2 | 74.37% ±0.85 | 71.67% ±0.89 | 2.70 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 55.83% ±2.11 | 43.29% ±2.31 | 12.53 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 76.57% ±1.92 | 86.88% ±1.94 | 10.31 pts | Reported ±1 SE ranges do not overlap |
| ProofBench v1.1 | 31.00% ±4.65 | 31.00% ±4.65 | 0.00 pts | Reported ±1 SE ranges overlap |
| GPQA Diamond | 92.68% ±1.46 | 92.93% ±1.29 | 0.25 pts | Reported ±1 SE ranges overlap |
| MMLU Pro | 89.52% ±0.31 | 89.22% ±0.31 | 0.30 pts | Reported ±1 SE ranges overlap |
| MMMU Pro | 88.27% ±0.77 | 61.79% ±1.17 | 26.47 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 49.88% ±3.44 | 34.97% ±3.32 | 14.92 pts | Reported ±1 SE ranges do not overlap |
| Code Migration | 26.75% ±4.10 | 36.60% ±4.14 | 9.85 pts | Reported ±1 SE ranges do not overlap |
| LiveCodeBench | 87.60% ±0.95 | 87.35% ±0.96 | 0.25 pts | Reported ±1 SE ranges overlap |
| ProgramBench | 0.00% ±0.00 | 0.00% ±0.00 | 0.00 pts | Reported ±1 SE ranges overlap |
| SkillsBench | 52.74% ±4.47 | 66.03% ±4.60 | 13.30 pts | Reported ±1 SE ranges do not overlap |
| SWE-bench | 78.80% ±1.83 | 86.60% ±1.52 | 7.80 pts | Reported ±1 SE ranges do not overlap |
| Terminal-Bench 4.0 | 6.06% ±1.51 | 8.59% ±1.34 | 2.53 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 48.68% ±4.73 | 69.00% ±4.48 | 20.32 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Gemini 3.5 Flash average | Grok 4.5 average |
|---|---|---|
| Legal | 38.96% | 45.62% |
| Finance | 63.48% | 59.81% |
| Healthcare | 66.20% | 65.09% |
| Math | 31.00% | 31.00% |
| Academic | 90.15% | 81.31% |
| Education | 49.88% | 34.97% |
| Coding | 42.95% | 50.60% |
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 3.5 Flash cost | Grok 4.5 cost | Gemini 3.5 Flash latency | Grok 4.5 latency |
|---|---|---|---|---|
| Vals Index | $3.30 | $2.46 | 18m17s | 15m44s |
| Harvey's Legal Agent Benchmark | $2.12 | $2.02 | 7m58s | 10m15s |
| Legal Research Bench | $1.31 | $1.09 | 4m38s | 15m54s |
| LegalBench | N/A | N/A | 34.58s | 67.88s |
| EMB | $5.17 | $1.55 | 18m44s | 18m28s |
| Finance Agent (v2) | $2.51 | $1.16 | 5m22s | 4m49s |
| MortgageTax | N/A | N/A | 11.43s | 117.46s |
| Tax Agent Bench | $0.86 | $0.47 | 3m53s | 4m44s |
| TaxEval v2 | N/A | N/A | 19.69s | 3m15s |
| MedCode | N/A | N/A | 25.29s | 59.64s |
| MedScribe | N/A | N/A | 57.08s | 11m45s |
| ProofBench v1.1 | $0.57 | $0.57 | 6m34s | 11m09s |
| GPQA Diamond | N/A | N/A | 27.66s | 3m38s |
| MMLU Pro | N/A | N/A | 26.26s | 5m58s |
| MMMU Pro | N/A | N/A | 12.23s | 11m55s |
| SAGE | N/A | N/A | 32.59s | 2m50s |
| Code Migration | $5.73 | $3.37 | 19m28s | 23m29s |
| LiveCodeBench | N/A | N/A | 64.37s | 9m49s |
| ProgramBench | N/A | N/A | 18m25s | 18m59s |
| SkillsBench | $1.25 | $0.58 | 7m15s | 2m16s |
| SWE-bench | $0.95 | $0.54 | 4m14s | 3m20s |
| Terminal-Bench 4.0 | $5.98 | $5.79 | 1h11m | 33m51s |
| Vibe Code Bench v1.1 | $2.54 | $4.10 | 14m53s | 13m50s |
Results available only for Gemini 3.5 Flash
- Terminal-Bench Science
- Public Benefits Bench v1.1
Results available only for Grok 4.5
- IOI
- CyberBench v1.1
- SRE Bench
- Time Horizon Index: KSP