Gemini 3 Pro (High): benchmark results
Gemini 3 Pro evaluated at the high reasoning-effort setting. Provider: Google. Released 2025-11-18. Access: API.
Aggregate Rating 1770 ± 14, rank #197 of 2168 rated models, from 86 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BaFCo - Coarse Layout Analysis (Zero-shot) | 26.46 | Mean average precision at IoU 0.3 (0-100): class-wise averag | 100 |
| BaFCo - Layout Analysis (CoT) | 11.34 | Mean average precision at IoU 0.3 (0-100): class-wise averag | 100 |
| BaFCo - Layout Analysis (Zero-shot) | 11.77 | Mean average precision at IoU 0.3 (0-100): class-wise averag | 100 |
| MM-JudgeBias - Detail Description | 0.94 | Bias-Conformity (0 to 1) on MM-JudgeBias (about 200 image-qu | 100 |
| MM-JudgeBias - Response Dominance | 0.99 | Bias-Deviation (0 to 1) on MM-JudgeBias (about 200 image-que | 100 |
| MM-JudgeBias - Text Dominance | 0.91 | Bias-Deviation (0 to 1) on MM-JudgeBias (about 200 image-que | 100 |
| VTC-Bench (Code Tools) | 51.18 | Average pass rate (%) answering VTC-Bench's 680 visual quest | 100 |
| VTC-Bench (Code Tools) - Attention Focusing | 62.22 | Pass rate (%) on attention focusing (45 questions on rotated | 100 |
| VTC-Bench (Code Tools) - Chart | 60 | Pass rate (%) on chart reasoning (100 questions), answering | 100 |
| VTC-Bench (Code Tools) - Color | 50 | Pass rate (%) on colour extraction (90 questions), answering | 100 |
| VTC-Bench (Code Tools) - Measurement | 54.29 | Pass rate (%) on measurement (105 questions on size, positio | 100 |
| VTC-Bench (Code Tools) - Robust OCR | 74 | Pass rate (%) on robust OCR (50 questions on degraded text), | 100 |
Interactive version: theaggregate.ai/model?slug=gemini-3-pro-high · How It Works · Data refreshed daily, snapshot 2026-10-20.