KazakhOCR (Cyrillic): leaderboard
Metric: Character error rate of the transcription of Cyrillic-script Kazakh (Wikipedia text) on 200 synthetic KazakhOCR images per script (random fonts, sizes, noise, blur, rotation and colours), the model transcribing the text zero-shot at temperature 0; error rate in percent, which can exceed 100 when the output adds text; lower is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 VL 7B Instruct | 5.3 | #643 |
| 2 | Llama 3.2 11B Instruct | 18.2 | #1112 |
| 3 | Gemma 3 12B (IT) | 25.7 | #655 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=kazakhocr-cyrillic · How It Works · Data refreshed daily, snapshot 2026-10-11.