Cedar - Text-Only - Arabic: leaderboard

Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around May 2028. 16 models tracked.

Top models

#ModelScore
1Gemini 2.5 Flash (Thinking)51.48
2Claude Sonnet 4.549.69
3Qwen 3 VL 235B A22B (Thinking)47.95
4GPT-4o (2024-11-20)46.56
5Kimi K244.15
6Qwen 3 VL 8B (Thinking)39.36
7Gemma 3 27B (IT)39.13
8aya-vision-8B36
9Pixtral-12B-240932.65
10Mistral Small 3.220.12

Interactive version: theaggregate.ai/benchmark?slug=cedar-text-only-arabic · How It Works · Data refreshed daily, snapshot 2026-09-25.