HELM Arabic - Mbzuai Human Translated Arabic Mmlu - College Physics: leaderboard
Metric: Exact Match (%). Source: crfm.stanford.edu. 41 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.7 | 93.14 |
| 2 | GPT-5.4 (2026-03-05) | 83.33 |
| 3 | Claude Haiku 4.5 (20251001) | 80.39 |
| 4 | Gemini 2.5 Flash (Thinking) | 79.41 |
| 5 | Mistral Large 3 | 78.43 |
| 6 | DeepSeek V3.1 | 76.47 |
| 7 | Gemma 4 31B (IT) | 75.49 |
| 8 | Llama 4 Maverick Instruct FP8 | 73.53 |
| 9 | Qwen 3.5 397B A17B | 70.59 |
| 10 | Claude Sonnet 4.6 | 69.61 |
| 11 | Gemini 2.5 Flash Lite (Thinking) | 61.76 |
| 12 | Llama 4 Scout Instruct | 60.78 |
| 13 | GPT-4.1 Mini | 59.8 |
| 14 | GPT-4.1 (2025-04-14) | 59.8 |
| 15 | Qwen 2.5 72B Instruct | 58.82 |
Interactive version: theaggregate.ai/benchmark?slug=helm-arabic-mbzuai-human-translated-arabic-mmlu-college-physics · How It Works · Data refreshed daily, snapshot 2026-09-19.