K-Bench: leaderboard
Metric: Overall (%). Source: arxiv.org. 30 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 98.96 |
| 2 | GPT-5.2 | 98.95 |
| 3 | Kimi K2.5 | 98.63 |
| 4 | Claude Opus 4.8 | 98.58 |
| 5 | Claude Fable 5 | 98.57 |
| 6 | GPT-5.4 | 98.51 |
| 7 | Claude Haiku 4.5 | 98.45 |
| 8 | Kimi K2.6 | 98.43 |
| 9 | Grok 4.20 | 98.32 |
| 10 | Gemini 3.1 Flash Lite | 98.13 |
| 11 | GPT-5.6 Terra | 98.05 |
| 12 | GPT-5.6 Sol | 97.98 |
| 13 | Nemotron 3 Ultra | 97.83 |
| 14 | Gemini 3.5 Flash | 97.71 |
| 15 | Qwen 3 32B | 97.65 |
Interactive version: theaggregate.ai/benchmark?slug=k-bench · How It Works · Data refreshed daily, snapshot 2026-09-20.