Gemini 3.1 Pro (Preview) (Medium): benchmark results
Provider: Google. Released 2026-02-19. Access: API.
Unified ELO 1719 ± 1, rank #91 of 3078 rated models, from 29 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Chess Bench LLM | 2010 | Lichess Rating | 99.2 |
| Context Arena MRCR (8-needle) | 48.5 | AUC@1M (%) | 97.7 |
| ObviousBench | 100 | Answer pass³ (%) | 97.7 |
| Conceptual Reasoning Index - Decision Theory (DTBench) | 95.12 | Chance-Corrected Score (0-100) | 95.6 |
| Context Arena MRCR (2-needle) | 87.7 | AUC@1M (%) | 95 |
| BeQu - Experiment 1 - Entailment Precision | 79.2 | Entailment Precision (%) | 94.7 |
| NOLLI - Jamo Composition | 57.3 | Accuracy (%) | 92.9 |
| GIM | 1.85 | IRT ability (theta) | 91.1 |
| Context Arena MRCR (4-needle) | 62.8 | AUC@1M (%) | 90.2 |
| OckBench | 87.5 | Accuracy (%) | 86.4 |
| NOLLI - Average | 73.6 | Accuracy (%; 25-task macro average) | 85.7 |
| NOLLI - Cipher - English | 84.7 | Accuracy (%) | 85.7 |
Interactive version: theaggregate.ai/model?slug=gemini-3-1-pro-preview-medium · How It Works · Data refreshed daily, snapshot 2026-09-19.