KOR-Bench - Logic: leaderboard
KOR-Bench questions under 25 newly written logic rules. Chat models, zero-shot. Higher is better. Source: arxiv.org/abs/2410.06526.
Metric: Accuracy (%; zero-shot, 250 questions). Source: arxiv.org. Saturation forecast: Estimated already saturated. 38 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet (20240620) | 67.2 | #309 |
| 2 | O1 Preview (2024-09-12) | 63.2 | #220 |
| 3 | O1 Mini (2024-09-12) | 61.2 | #414 |
| 4 | Llama 3.1 405B Instruct | 56.8 | #472 |
| 5 | Qwen 2.5 32B Instruct | 56.8 | #511 |
| 6 | GPT-4 Turbo | 54 | #419 |
| 7 | Qwen 2.5 72B Instruct | 53.2 | #470 |
| 8 | GPT-4o (2024-05-13) | 52.4 | #338 |
| 9 | Mistral Large 2 (Jul) | 51.2 | #498 |
| 10 | Qwen 2.5 14B Instruct | 50 | #657 |
| 11 | Llama 3.1 70B Instruct | 49.2 | #570 |
| 12 | Gemma 2 27B (IT) | 49.2 | #655 |
| 13 | DeepSeek V2.5 | 48 | #503 |
| 14 | Yi Large | 47.6 | #565 |
| 15 | Llama 3 70B Instruct | 46.4 | #643 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #6 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=kor-bench-logic · How It Works · Data refreshed daily, snapshot 2026-11-07.