Gemini 3.1 Pro Preview scores 41.4 on the Respan Index (#42), ahead of Gemini 3.6 Flash at 40.9 (#43). Gemini 3.1 Pro Preview leads on 15 of the 24 benchmarks both report. Gemini 3.6 Flash is 3.0x cheaper per token at list price.
* Estimated: no published scores in that category. How the Respan Index works
| FrontierMath Tiers 1-3 v2 (Epoch AI) |
| 59.6% |
| 58.9% |
| Furniture Assembly (Epoch AI) | 26.7% | 23.3% |
| GPQA Diamond | 94.3% | 94.1% |
| LiveBench Agentic Coding (LiveBench) | 44.1% | 43.4% |
| LiveBench Coding (LiveBench) | 76.5% | 77.9% |
| LiveBench Data Analysis (LiveBench) | 78.5% | 63% |
| LiveBench Instruction Following (LiveBench) | 79.1% | 75.4% |
| LiveBench Language (LiveBench) | 85.4% | 83.9% |
| LiveBench Mathematics (LiveBench) | 91% | 86.4% |
| LiveBench Reasoning (LiveBench) | 84% | 85.1% |
| Mystery Game Puzzles (Epoch AI) | 34% | 30% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 95.6% | 94.2% |
| SAGE (Vals AI) | 48.7% | 48.7% |
| SimpleQA Verified | 73.5% | 66.2% |
| WeirdML (Håvard Tveit Ihle) | 72.1% | 56.1% |
| LMArena Agent (LMArena) | -0.0771 | -0.0768 |
| LMArena Elo | 1480 | 1483 |
| LMArena Vision (LMArena) | 1279 | 1280 |
| LMArena WebDev (LMArena) | 1446 | 1536 |
| AIME 2026 | 98.3% | 96.7% |
Higher is better. Each score links to where it was published.