HELM Arabic - Alghafa - Multiple Choice Grounded Statement Soqal Task: leaderboard
Metric: Exact Match (%). Source: crfm.stanford.edu. 41 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 | 96.67 |
| 2 | Gemma 4 31B (IT) | 96.67 |
| 3 | Mistral Large 3 | 96.67 |
| 4 | Gemini 2.5 Flash (Thinking) | 96 |
| 5 | Gemini 2.5 Flash Lite (Thinking) | 96 |
| 6 | GPT-4.1 (2025-04-14) | 95.33 |
| 7 | Llama 4 Maverick Instruct FP8 | 95.33 |
| 8 | DeepSeek V3.1 | 95.33 |
| 9 | Mistral Large 2 (Nov) Instruct (2411) | 95.33 |
| 10 | GPT-5.4 (2026-03-05) | 95.33 |
| 11 | Llama 3.3 70B Instruct | 94.67 |
| 12 | Qwen 2.5 72B Instruct | 94 |
| 13 | Llama 4 Scout Instruct | 94 |
| 14 | Claude Opus 4.7 | 94 |
| 15 | Claude Haiku 4.5 (20251001) | 94 |
Interactive version: theaggregate.ai/benchmark?slug=helm-arabic-alghafa-multiple-choice-grounded-statement-soqal-task · How It Works · Data refreshed daily, snapshot 2026-09-19.