Gemini 3 Pro (Preview) (Low): benchmark results
Provider: Google. Released 2025-11-18. Access: API.
Aggregate Rating 1743 ± 19, rank #251 of 2169 rated models, from 64 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| VisReason (Vision-Centric) | 47.5 | Average accuracy (%) over the 10 reasoning categories (1,505 | 100 |
| VisReason (Vision-Centric) - 3D-Spatial Reasoning | 43.9 | Accuracy (%) on the 3D-Spatial Reasoning category (structura | 100 |
| VisReason (Vision-Centric) - Board Reasoning | 51.1 | Accuracy (%) on the Board Reasoning category (structural rea | 100 |
| VisReason (Vision-Centric) - Bounding-Box | 24.5 | Accuracy (%) on the 140 bounding-box questions (boxes Hungar | 100 |
| VisReason (Vision-Centric) - Cue Insight | 68.5 | Accuracy (%) on the Cue Insight category (conceptual reasoni | 100 |
| VisReason (Vision-Centric) - Deductive Reasoning | 53.7 | Accuracy (%) on the Deductive Reasoning category (conceptual | 100 |
| VisReason (Vision-Centric) - Fill-in-the-Blank | 47.4 | Accuracy (%) on the 441 fill-in-the-blank questions (answer | 100 |
| VisReason (Vision-Centric) - Geolocation | 76 | Accuracy (%) on the Geolocation category (conceptual reasoni | 100 |
| VisReason (Vision-Centric) - Inductive Reasoning | 53.1 | Accuracy (%) on the Inductive Reasoning category (conceptual | 100 |
| VisReason (Vision-Centric) - Localized Reasoning | 44.2 | Accuracy (%) on the Localized Reasoning category (perceptual | 100 |
| VisReason (Vision-Centric) - Multiple-Choice | 47.8 | Accuracy (%) on the 615 multiple-choice questions (option in | 100 |
| VisReason (Vision-Centric) - Open-Ended | 72.5 | Accuracy (%) on the 309 open-ended questions (judged correct | 100 |
Interactive version: theaggregate.ai/model?slug=gemini-3-pro-preview-low · How It Works · Data refreshed daily, snapshot 2026-10-23.