LegalBench ↗87.03 %
Gemini 3 Pro (11/25) · LegalBench · Accuracy ↗
Google / MODEL EVIDENCE
Gemini 3 Pro (11/25)
18 benchmarks · 19 observations · 19 sources
Scores at a glance
18 benchmarks · large dot: Gemini 3 Pro (11/25); small dots: closest models where reported
CaseLaw v2 ↗53.05 %
Gemini 3 Pro (11/25) · CaseLaw · Accuracy ↗
CorpFin v2 ↗63.67 %
Gemini 3 Pro (11/25) · CorpFin · Accuracy ↗
MortgageTax & TaxEval v2 ↗69.08 %
Gemini 3 Pro (11/25) · MortgageTax · Accuracy ↗
1 more reported result
72.57 % · Gemini 3 Pro (11/25) · TaxEval · Accuracy · Vals AI · TaxEval ↗
MedCode ↗52.2 %
Gemini 3 Pro (11/25) · MedCode · Accuracy ↗
MedScribe ↗72.04 %
Gemini 3 Pro (11/25) · MedScribe · Accuracy ↗
MedQA (USMLE) ↗96.03 %
Gemini 3 Pro (11/25) · MedQA · Accuracy ↗
AIME 2024–2025 · Vals ↗96.68 %
Gemini 3 Pro (11/25) · AIME · Accuracy ↗
MATH-500 ↗96.4 %
Gemini 3 Pro (11/25) · MATH 500 · Accuracy ↗
MGSM (Multilingual GSM) ↗93.93 %
Gemini 3 Pro (11/25) · MGSM · Accuracy ↗
GPQA Diamond ↗91.67 %
Gemini 3 Pro (11/25) · GPQA Diamond · Accuracy ↗
MMLU-Pro ↗90.1 %
Gemini 3 Pro (11/25) · MMLU Pro · Accuracy ↗
MMMU-Pro ↗87.51 %
Gemini 3 Pro (11/25) · MMMU Pro · Accuracy ↗
Gemini 3 Pro (11/25) · SAGE · Accuracy ↗
Vibe Code Bench ↗14.3 %
Gemini 3 Pro (11/25) (OpenHands) · Vibe Code Bench · original task set ↗
SWE-bench Verified ↗76.4 %
Gemini 3 Pro (11/25) (Mini-SWE-agent) · SWE-bench Verified · Score ↗
LiveCodeBench ↗86.41 %
Gemini 3 Pro (11/25) · LiveCodeBench · Accuracy ↗
Poker Agent Benchmark ↗1078.9 TrueSkill
Gemini 3 Pro (11/25) · Poker Agent · TrueSkill Rating ↗
Closest models by score
Similar positions on shared benchmark leaderboards. Smaller gaps are closer.
Gemini 3 Flash (12/25) →Google
15 of 18 shared benchmark comparisons
13 of 18 shared benchmark comparisons
GPT-5.4 →OpenAI
13 of 18 shared benchmark comparisons
How similarity is measured
Mean absolute gap in percentile positions on shared benchmark settings; lower is closer. At least 60% overlap is required, with a minimum of three when available. Independent evaluations are used when there are at least three; otherwise published report comparisons are used. Repeated model observations within one setting are averaged before ranking. Writing-style traits and composite indices are excluded. This describes similarity, not an overall quality ranking.
Sources
Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.
- AIME — evaluation methodology and results ↗
- CaseLaw — evaluation methodology and results ↗
- CorpFin — evaluation methodology and results ↗
- GPQA Diamond — evaluation methodology and results ↗
- LiveCodeBench — evaluation methodology and results ↗
- LegalBench — evaluation methodology and results ↗
- MATH 500 — evaluation methodology and results ↗
- MedCode — evaluation methodology and results ↗
- MedQA — evaluation methodology and results ↗
- MedScribe — evaluation methodology and results ↗
- MGSM — evaluation methodology and results ↗
- MMLU Pro — evaluation methodology and results ↗
- MMMU Pro — evaluation methodology and results ↗
- MortgageTax — evaluation methodology and results ↗
- Poker Agent — evaluation methodology and results ↗
- SAGE — evaluation methodology and results ↗
- SWE-bench Verified — evaluation methodology and results ↗
- TaxEval — evaluation methodology and results ↗
- Vibe Code Bench — evaluation methodology and results ↗
Names used in source tables
Gemini 3 Pro (11/25) · Gemini 3 Pro (11/25) (Mini-SWE-agent) · Gemini 3 Pro (11/25) (OpenHands)