KRETA - System 1: leaderboard
The 1,426 KRETA items that only require reading and understanding the Korean text in the image; the score is the share answered correctly. It isolates text recognition from the reasoning items of System 2.
Metric: Accuracy on 1,426 System 1 (text recognition and understanding) items (%). Source: github.com. Saturation forecast: Estimated already saturated. 17 models tracked. Updated 17 Nov 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o (2024-11-20) | 95.9 | #864 |
| 2 | A.X-4.0-VL-Light | 95.3 | #2575 |
| 3 | Qwen 2.5 VL 7B Instruct | 94.5 | #1893 |
| 4 | Qwen 2.5 VL 3B Instruct | 94.2 | #2648 |
| 5 | kanana-1.5-v-3B-instruct | 94 | #3953 |
| 6 | VARCO-VISION-2.0-14B | 93.5 | #1794 |
| 7 | Claude 3.5 Sonnet (20241022) | 93.4 | #648 |
| 8 | VARCO-VISION-14B | 90.9 | #1854 |
| 9 | InternVL2.5-4B | 90.7 | #2559 |
| 10 | InternVL2.5-8B | 89.8 | #2401 |
| 11 | MiniCPM-o-2.6 | 84.1 | #2417 |
| 12 | Ovis1.6-Gemma2-9B | 68.9 | #2365 |
| 13 | DeepSeek-VL2-Small | 67.3 | #2502 |
| 14 | Ovis1.6-Llama3.2-3B | 62.8 | #3750 |
| 15 | DeepSeek-VL2-Tiny | 60.8 | #6676 |
No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=kreta-system-1 · How It Works · Data refreshed daily, snapshot 2026-11-17.