HELM Arabic - Mbzuai Human Translated Arabic Mmlu - World Religions: leaderboard

Metric: Exact Match (%). Source: crfm.stanford.edu. 41 models tracked.

Top models

#ModelScore
1Claude Sonnet 4.690.64
2Claude Opus 4.790.64
3GPT-5.4 (2026-03-05)89.47
4Gemma 4 31B (IT)87.72
5Gemini 2.5 Flash (Thinking)87.72
6Mistral Large 385.96
7GPT-4.1 (2025-04-14)84.21
8DeepSeek V3.184.21
9Qwen 3.5 397B A17B82.46
10Qwen 2.5 72B Instruct81.29
11Claude Haiku 4.5 (20251001)80.7
12AceGPT-v2-70B-Chat79.53
13Llama 3.3 70B Instruct78.95
14Mistral Large 2 (Nov) Instruct (2411)78.95
15Gemini 2.5 Flash Lite (Thinking)78.95

Interactive version: theaggregate.ai/benchmark?slug=helm-arabic-mbzuai-human-translated-arabic-mmlu-world-religions · How It Works · Data refreshed daily, snapshot 2026-09-19.