OpenCompass Knowledge - Humanities - English (CompassBench 2411): leaderboard

Metric: Score (%). Source: rank.opencompass.org.cn. 34 models tracked.

Top models

#ModelScore
1GPT-4o (2024-11-20)84.2
2GPT-4o (2024-08-06)83.5
3Claude 3.5 Sonnet (20241022)79.9
4Step 2 16K79.9
5Llama 3.1 405B Instruct FP878.4
6Llama 3.1 70B Instruct76.3
7Mistral Large 2 (Nov) Instruct (2411)76.3
8GLM-4 Plus76.3
9Qwen 2.5 72B Instruct73.4
10Grok Beta73.4
11Yi Lightning71.9
12DeepSeek V2.571.2
13TeleChat268.3
14Mistral-Small-Instruct-240964
15O1 Mini (2024-09-12)63.3

Interactive version: theaggregate.ai/benchmark?slug=opencompass-knowledge-humanities-english-compassbench-2411 · How It Works · Data refreshed daily, snapshot 2026-09-19.