CZ-EVAL - Critical Thinking — leaderboard

Metric: Accuracy (%). Source: huggingface.co. 12 models tracked.

Top models

#ModelScore
1Claude 3 Opus76.45
2GPT-4 Preview (1106)70.65
3Claude 3 Sonnet68.89
4Mistral Large66.78
5Claude 2.164.5
6Gemini 1.0 Pro60.98
7Claude 3 Haiku60.46
8Mistral Medium59.58
9Command R53.43
10Mixtral 8x7B Instruct51.14
11GPT-3.5 Turbo49.91
12gemma-7B (IT)27.94

Interactive version: aibenchmarks.dev/benchmark?slug=cz-eval-critical-thinking · How the rankings work · Data refreshed daily, snapshot 2026-07-08.