HELM Arabic Enterprise - Arabic Scenarios: leaderboard
Metric: Mean score. Source: crfm.stanford.edu. 35 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.7 | 77.91 |
| 2 | GPT-5.4 (2026-03-05) | 76.68 |
| 3 | Gemma 4 31B (IT) | 73.8 |
| 4 | Claude Sonnet 4.6 | 73.67 |
| 5 | DeepSeek V4 Pro | 72.84 |
| 6 | Qwen3 235B A22B Instruct 2507 FP8 | 69.59 |
| 7 | Gemini 2.5 Flash (Thinking) | 68.17 |
| 8 | Mistral Large 3 | 66.02 |
| 9 | DeepSeek V3.1 | 65.92 |
| 10 | Qwen 3.5 397B A17B | 65.27 |
| 11 | Llama 4 Maverick Instruct | 64.83 |
| 12 | Claude Haiku 4.5 (20251001) | 63.63 |
| 13 | Gemini 2.5 Flash Lite (Thinking) | 62.6 |
| 14 | Qwen 3.5 9B | 62.42 |
| 15 | Llama 4 Scout Instruct | 62.07 |
Interactive version: theaggregate.ai/benchmark?slug=helm-arabic-enterprise-arabic-scenarios · How It Works · Data refreshed daily, snapshot 2026-09-08.