Gemini 3.8 Flash scores 52.2 on the Respan Index (#16), ahead of Kimi K3 at 50.6 (#21). Gemini 3.8 Flash leads on 16 of the 28 benchmarks both report. Gemini 3.8 Flash is 4.0x cheaper per token at list price.
* Estimated: no published scores in that category. How the Respan Index works
| 68.4% |
| 72.2% |
| FrontierSWE V2 (Proximal Labs) | 19.6% | 25.9% |
| Furniture Assembly (Epoch AI) | 31.7% | 34.2% |
| GPQA Diamond | 95.3% | 93.5% |
| LiveBench Agentic Coding (LiveBench) | 54.2% | 62.2% |
| LiveBench Coding (LiveBench) | 72.5% | 81.4% |
| LiveBench Data Analysis (LiveBench) | 54% | 78.7% |
| LiveBench Instruction Following (LiveBench) | 81.4% | 71.4% |
| LiveBench Language (LiveBench) | 87.8% | 85.5% |
| LiveBench Mathematics (LiveBench) | 91.6% | 84.4% |
| LiveBench Reasoning (LiveBench) | 89.3% | 90.7% |
| Mystery Game Puzzles (Epoch AI) | 47% | 26% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 98.9% | 97.2% |
| SAGE (Vals AI) | 35.1% | 52.8% |
| SimpleQA Verified | 69.7% | 50.6% |
| Terminal-Bench 4.0 (Vals AI) | 19.2% | 17.2% |
| Vending-Bench 2 (Andon Labs) | 5093.79 | 5165.04 |
| WeirdML (Håvard Tveit Ihle) | 84.8% | 82.6% |
| LMArena Agent (LMArena) | 0.0297 | 0.0418 |
| LMArena Elo | 1495 | 1488 |
| LMArena WebDev (LMArena) | 1583 | 1658 |
| SimpleBench (SimpleBench) | 82.4% | 60.7% |
| CharXiv Reasoning (no tools) | 86.2% | 84.8% |
| DeepSWE v1.1 | 73.7% | 67.3% |
Higher is better. Each score links to where it was published.