Cedar - Text-Only - English: leaderboard

Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around August 2028. 17 models tracked.

Top models

#ModelScore
1Claude Sonnet 4.561.79
2Gemini 2.5 Flash (Thinking)57.53
3Kimi K256.72
4Qwen 3 VL 235B A22B (Thinking)55.15
5Gemma 3 27B (IT)53.91
6GPT-4o (2024-11-20)53.69
7Qwen 3 VL 8B (Thinking)49.7
8aya-vision-8B45.96
9Llama 3.2 11B Instruct43.04
10Pixtral-12B-240942.37
11Mistral Small 3.239.97

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