Kimi K3 scores 50.6 on the Respan Index (#21), ahead of Qwen3.8-Max at 50.3 (#22). Kimi K3 leads on 13 of the 25 benchmarks both report. Qwen3.8-Max is 2.0x cheaper per token at list price. Kimi K3 has the larger context window (1.0M tokens).
* Estimated: no published scores in that category. How the Respan Index works
| 72.2% |
| 74.7% |
| FrontierSWE V2 (Proximal Labs) | 25.9% | 17.8% |
| Furniture Assembly (Epoch AI) | 34.2% | 20% |
| GPQA Diamond | 93.5% | 92.6% |
| LiveBench Agentic Coding (LiveBench) | 62.2% | 64.6% |
| LiveBench Coding (LiveBench) | 81.4% | 72.9% |
| LiveBench Data Analysis (LiveBench) | 78.7% | 78.4% |
| LiveBench Instruction Following (LiveBench) | 71.4% | 74.1% |
| LiveBench Language (LiveBench) | 85.5% | 79.7% |
| LiveBench Mathematics (LiveBench) | 84.4% | 91.3% |
| LiveBench Reasoning (LiveBench) | 90.7% | 88.2% |
| Mystery Game Puzzles (Epoch AI) | 26% | 38% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 97.2% | 99.4% |
| SAGE (Vals AI) | 52.8% | 51.3% |
| SimpleQA Verified | 50.6% | 45.8% |
| Terminal-Bench 4.0 (Vals AI) | 17.2% | 34.3% |
| LMArena Agent (LMArena) | 0.0418 | 0.0251 |
| LMArena Elo | 1488 | 1481 |
| LMArena WebDev (LMArena) | 1658 | 1671 |
| tau2-bench Banking Knowledge (Sierra) | 37.1% | 55.2% |
| Terminal-Bench 2.1 | 88.3% | 86.6% |
Higher is better. Each score links to where it was published.