Kimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI. It is suited for complex coding, knowledge work, and long-horizon agentic workflows, and is particularly strong at... (description from the OpenRouter listing)
| Benchmark | Domain | Score | vs best recorded | Setting | Run | Source |
|---|---|---|---|---|---|---|
| Furniture Assembly | multimodal | 34.2% ±6.1 | 41% | max | 11 Sep 2026 | Epoch ↗ |
| LMCA | agents | 52.7% | 81% | max | — | External ↗ |
| DTBench | reasoning | 91.2% | 93% | max | — | External ↗ |
| Mystery Game Puzzles | games | 26.0% ±4.4 | 31% | max | 29 Jul 2026 | Epoch ↗ |
| Surface Evolver Bench | science | 93.0% | 98% | max | — | External ↗ |
| FrontierMath-Tiers-1-3-v2-Private | math | 72.2% ±2.7 | 77% | max | 17 Jul 2026 | Epoch ↗ |
| FrontierMath-Tier-4-v2-Private | math | 39.0% ±7.7 | 40% | max | 17 Jul 2026 | Eval log ↗ |
| DeepSWE | coding | 68.5% | 92% | max | — | External ↗ |
| Chess Puzzles | games | 39.0% ±4.9 | 54% | max | 16 Jul 2026 | Epoch ↗ |
| SimpleQA Verified | knowledge | 50.6% ±1.6 | 67% | max | 27 Aug 2026 | Eval log ↗ |
| ARC-AGI-2 | reasoning | 60.4% | 64% | max | — | External ↗ |
| WeirdML | coding | 82.6% ±0.0 | 88% | max | — | External ↗ |
| OTIS Mock AIME 2024-2025 | math | 97.2% ±1.1 | 97% | max | 16 Jul 2026 | Epoch ↗ |
| SimpleBench | reasoning | 60.7% | 74% | max | — | External ↗ |
| GPQA diamond | science | 93.1% ±1.5 | 97% | max | 16 Jul 2026 | Epoch ↗ |
| ARC-AGI | reasoning | 94.5% | 96% | max | — | External ↗ |
Source: Epoch AI Benchmarking Hub (CC BY 4.0). “External” rows are leaderboard results Epoch collects from third parties. Best reported setting per benchmark is shown.
4 observations (OpenRouter listing + Internet Archive snapshots). History accumulates with every ingest run; a single point means no change has been observed yet.