HELM Arabic - Mbzuai Human Translated Arabic Mmlu - Global Facts: leaderboard
HELM Arabic (release v2.2.0) MBZUAI human-translated Arabic MMLU, Global Facts: the MMLU four-option questions for this subject translated into Arabic by human translators; 100 questions, exact match. Percent correct, 0-100, higher is better; guessing scores 25.
Metric: Exact Match (%). Source: crfm.stanford.edu. Saturation forecast: Around February 2027. 41 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 | 71 | #51 |
| 2 | Claude Sonnet 4.6 | 67 | #94 |
| 3 | Gemma 4 31B (IT) | 67 | #178 |
| 4 | Mistral Large 3 | 67 | #410 |
| 5 | GPT-5.1 (2025-11-13) | 67 | #107 |
| 6 | DeepSeek V3.1 | 66 | #275 |
| 7 | Mistral Large 2 (Nov) Instruct (2411) | 65 | #480 |
| 8 | GPT-4.1 (2025-04-14) | 62 | #257 |
| 9 | Gemini 2.5 Flash (Non-reasoning) | 62 | #255 (Gemini 2.5 Flash) |
| 10 | GPT-5.4 (2026-03-05) | 60 | #63 |
| 11 | Llama 3.3 70B Instruct | 59 | #541 |
| 12 | Qwen 3.5 397B A17B | 59 | #154 |
| 13 | Qwen 3 Next 80B A3B Instruct | 58 | #381 |
| 14 | Claude Haiku 4.5 (20251001) | 55 | #264 |
| 15 | Llama 4 Maverick Instruct FP8 | 55 | #424 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-arabic-mbzuai-human-translated-arabic-mmlu-global-facts · How It Works · Data refreshed daily, snapshot 2026-11-05.