HELM Arabic - Arabic Mmlu - Social Science (Middle School): leaderboard
HELM Arabic (release v2.2.0) ArabicMMLU Social Science (Middle School): native Arabic multiple-choice questions from ArabicMMLU, which draws on school, university and professional exams across Arabic-speaking countries; 241 questions, exact match. Percent correct, 0-100, higher is better.
Metric: Exact Match (%). Source: crfm.stanford.edu. Saturation forecast: Around June 2028. 41 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 | 64.32 | #51 |
| 2 | GPT-5.4 (2026-03-05) | 62.66 | #63 |
| 3 | Llama 3.3 70B Instruct | 62.24 | #541 |
| 4 | Mistral Large 3 | 62.24 | #410 |
| 5 | GPT-4.1 (2025-04-14) | 61.83 | #257 |
| 6 | GPT-5.1 (2025-11-13) | 61.83 | #107 |
| 7 | Gemma 4 31B (IT) | 61.41 | #178 |
| 8 | Llama 4 Maverick Instruct FP8 | 61.41 | #424 |
| 9 | Gemini 2.5 Flash (Non-reasoning) | 61.41 | #255 (Gemini 2.5 Flash) |
| 10 | Llama 4 Scout Instruct | 61 | #547 |
| 11 | Qwen 3.5 397B A17B | 60.58 | #154 |
| 12 | Qwen 3 Next 80B A3B Instruct | 60.58 | #381 |
| 13 | Claude Sonnet 4.6 | 60.17 | #94 |
| 14 | Qwen 2.5 72B Instruct | 59.34 | #469 |
| 15 | DeepSeek V3.1 | 59.34 | #275 |
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-arabic-mmlu-social-science-middle-school · How It Works · Data refreshed daily, snapshot 2026-11-05.