Gemini 3 Pro (Preview) (High): benchmark results

Gemini 3 Pro (Preview) evaluated at the high reasoning-effort setting. Provider: Google. Released 2025-11-18. Access: API.

Unified ELO 1717 ± 1, rank #68 of 1761 rated models, from 87 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AA LiveCodeBench91.75Pass@1 (%)100
AA MMLU-Pro89.76Accuracy (%)100
AA Omniscience - Software Engineering (SWE) - Java55Accuracy (%)100
Eleusis Benchmark17.26Avg Floored Score100
Vals AI MATH 50096.4Accuracy (%)100
AA Omniscience - Software Engineering (SWE) - JavaScript75.45Accuracy (%)99.7
AA Omniscience - Software Engineering (SWE) - Kotlin66Accuracy (%)99.7
AA Omniscience - Software Engineering (SWE) - Julia56Accuracy (%)99.5
AA Omniscience - Software Engineering (SWE) - Python66.5Accuracy (%)99.3
AA Omniscience - Law65.1Accuracy (%)99.2
AA Global-MMLU-Lite92.22Accuracy (%)99.1
AA Global-MMLU-Lite - Arabic92.67Accuracy (%)99.1

Interactive version: theaggregate.ai/model?slug=gemini-3-pro-preview-high · How It Works · Data refreshed daily, snapshot 2026-09-05.