SEAL - Arabic: leaderboard
Arabic-language tasks spanning practical assistance, writing, reasoning, and culturally situated scenarios, measuring fluency, cultural appropriateness, and factual accuracy in Arabic.
Metric: Score. Source: scale.com. Saturation forecast: Estimated already saturated. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O1 (2024-12-17) | 1120 | #149 |
| 2 | Gemini 2.0 Flash (Thinking) | 1120 | #342 (Gemini 2.0 Flash) |
| 3 | Gemini 1.5 Pro | 1116 | #445 |
| 4 | O3 Mini | 1093 | #272 |
| 5 | Gemini 2.0 Flash | 1090 | #342 |
| 6 | O1 Preview | 1087 | #193 |
| 7 | GPT-4o (2024-08-06) | 1066 | #336 |
| 8 | aya-expanse-32B | 1025 | #692 |
| 9 | GPT-4 | 1011 | #438 |
| 10 | Claude 3.5 Sonnet (20240620) | 995 | #299 |
| 11 | Mistral Large 2 (Jul) | 970 | #484 |
| 12 | Gemini 1.5 Flash | 967 | #618 |
| 13 | aya-23-35B | 932 | #883 |
| 14 | Llama 3.1 405B Instruct | 875 | #457 |
| 15 | Llama 3.3 70B Instruct | 808 | #534 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=seal-arabic · How It Works · Data refreshed daily, snapshot 2026-10-20.