Gemini 3.1 Pro (Preview) (Low) — benchmark results
Gemini 3.1 Pro (Preview) evaluated at the low reasoning-effort setting. Provider: Google. Released 2026-02-19. Access: API.
Unified ELO 1936 ± 57, rank #46 of 1776 rated models, from 9 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Context Arena MRCR (4-needle) | 63.6 | AUC@1M (%) | 95.1 |
| Context Arena MRCR (2-needle) | 87.4 | AUC@1M (%) | 93.8 |
| Context Arena MRCR (8-needle) | 47 | AUC@1M (%) | 93.2 |
| LisanBench | 0.24 | Mean Path Length / Current Maximum | 88.6 |
| ChessImageBench | 20.8 | All Correct (%) | 80 |
| GIM | 1.45 | IRT ability (theta) | 77.8 |
| NonoBench | 40 | Overall Accuracy (%) | 74.5 |
| MathArena - ARXIV February | 40.62 | Accuracy (%) | 52 |
| MathArena - ArXiv Math Jan 2026 | 50 | Accuracy (%) | 25 |
Interactive version: theaggregate.ai/model?slug=gemini-3-1-pro-preview-low · How the rankings work · Data refreshed daily, snapshot 2026-07-22.