KITAB-Bench
MBZUAI
8,809 Arabic text samples across 9 domains. Tests Arabic script recognition.
Benchmark Stats
Models8
Papers8
Metrics1
SOTA History
Not enough data to show trend.
Character Error Rate
Levenshtein distance between predicted and ground truth (lower is better)
Lower is better
| Rank | Model | Source | Score | Year | Paper |
|---|---|---|---|---|---|
| 1 | gemini-20-flash Arabic OCR - Character Error Rate (lower is better). 8,809 samples, 9 domains | Editorial | 0.13 | 2025 | Source |
| 2 | ain-7b | Editorial | 0.2 | 2025 | Source |
| 3 | gpt-4o | Editorial | 0.31 | 2025 | Source |
| 4 | gpt-4o-mini | Editorial | 0.43 | 2025 | Source |
| 5 | azure-ocr | Editorial | 0.52 | 2025 | Source |
| 6 | tesseract | Editorial | 0.54 | 2025 | Source |
| 7 | easyocr | Editorial | 0.58 | 2025 | Source |
| 8 | paddleocr | Editorial | 0.79 | 2025 | Source |