Who leads the KITAB-Bench benchmark?

gemini-20-flash currently leads KITAB-Bench with a score of 0.13 on Character Error Rate (lower is better).

What is the state-of-the-art score on KITAB-Bench?

The state-of-the-art result on KITAB-Bench is 0.13 (Character Error Rate), achieved by gemini-20-flash as of 2026.

Codesota tracks 8 models on KITAB-Bench.

The KITAB-Bench leaderboard on Codesota includes results through 2026.

Codesota · Benchmark · KITAB-BenchHome/Leaderboards/Vision & Documents/Document OCR/KITAB-Bench

MBZUAI

Name: KITAB-Bench Benchmark Results
Creator: MBZUAI
Published: 2026-01-01
License: https://creativecommons.org/licenses/by/4.0/

8,809 Arabic text samples across 9 domains. Tests Arabic script recognition.

§ 01 · Leaderboard

Found a wrong score or missing run?

Use row edits to send a sourced correction into moderation.

Levenshtein distance between predicted and ground truth (lower is better)

Lower is better

Trust tiers for Character Error Rateverifiedpapervendorcommunityunverified

Muted rows were not state of the art when published — an earlier or same-year result already scored better.

Rank	Model	Trust	Score	Year	Links	Fix
01	gemini-20-flash Non-API entry from src	unverified	0.13	2026	N/A	Looks wrong?
02	ain-7b Non-API entry from src	unverified	0.20	2026	N/A	Looks wrong?
03	gpt-4o Non-API entry from src	unverified	0.31	2026	N/A	Looks wrong?
04	gpt-4o-mini Non-API entry from src	unverified	0.43	2026	N/A	Looks wrong?
05	azure-ocr Non-API entry from src	unverified	0.52	2026	N/A	Looks wrong?
06	tesseract Non-API entry from src	unverified	0.54	2026	N/A	Looks wrong?
07	easyocr Non-API entry from src	unverified	0.58	2026	N/A	Looks wrong?
08	paddleocr Non-API entry from src	unverified	0.79	2026	N/A	Looks wrong?