# Scale Labs SEAL Arabic (SEAL Arabic)

> Scale's SEAL leaderboard for Arabic: expert raters judge how well models answer about 1,000 prompts written by native Arabic speakers. BenchLM shows it as display-only reference data, excluded from rankings. Scale marked this board deprecated as of March 2025 and no longer adds models.

Canonical page: https://benchlm.ai/benchmarks/scale-arabic

- Category: [external](/external)
- Last updated: November 12, 2026 snapshot

## About SEAL Arabic

- Year: 2026
- Tasks: About 1,000 prompts per language
- Format: Elo-style rating
- Difficulty: External agent and model evaluation
- Paper: [Scale Labs leaderboard](https://labs.scale.com/leaderboard/arabic)

Single-turn prompts across general and culturally specific scenarios, rated by native-speaking experts. Scores are ratings on an Elo-style scale built from expert human comparisons, so only the gaps between rows carry meaning. BenchLM mirrors 19 published rows captured on November 12, 2026 snapshot.

SEAL Arabic is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (19 models)

| Rank | Model | Creator | Score |
|------|-------|---------|-------|
| 1 | [Gemini 1.5 Pro (August 27, 2024)](https://labs.scale.com/leaderboard/arabic) | Google | 1147 |
| 2 | [Gemini 2.0 Pro (December 2024)](https://labs.scale.com/leaderboard/arabic) | Google | 1138 |
| 3 | [o1 (December 2024)](https://labs.scale.com/leaderboard/arabic) | OpenAI | 1120 |
| 4 | [Gemini 2.0 Flash Thinking (January 2025)](https://labs.scale.com/leaderboard/arabic) | Google | 1120 |
| 5 | [Gemini 1.5 Pro (November 2024)](https://labs.scale.com/leaderboard/arabic) | Google | 1116 |
| 6 | [o3-mini](/models/o3-mini) | OpenAI | 1093 |
| 7 | [Gemini Pro Flash 2](https://labs.scale.com/leaderboard/arabic) | Google | 1090 |
| 8 | [o1-preview](/models/o1-preview) | OpenAI | 1087 |
| 9 | [GPT-4o (August 2024)](https://labs.scale.com/leaderboard/arabic) | OpenAI | 1066 |
| 10 | [Aya Expanse 32B](https://labs.scale.com/leaderboard/arabic) | Cohere | 1025 |
| 11 | [GPT-4 (November 2024)](https://labs.scale.com/leaderboard/arabic) | OpenAI | 1011 |
| 12 | [Claude 3.5 Sonnet (June 2024)](https://labs.scale.com/leaderboard/arabic) | Anthropic | 995 |
| 13 | [Mistral Large 2](/models/mistral-large-2) | Mistral | 970 |
| 14 | [Gemini 1.5 Flash](/models/gemini-1-5-flash) | Google | 967 |
| 15 | [Aya 23 35B*](https://labs.scale.com/leaderboard/arabic) | Cohere | 932 |
| 16 | [Llama 3.1 405B Instruct](https://labs.scale.com/leaderboard/arabic) | Meta | 875 |
| 17 | [Llama 3.3 70B Instruct](https://labs.scale.com/leaderboard/arabic) | Meta | 808 |
| 18 | [Jais Adapted 70B](https://labs.scale.com/leaderboard/arabic) | Unknown | 787 |
| 19 | [Gemma 2 27B](/models/gemma-2-27b) | Google | 661 |

## FAQ

### What does SEAL Arabic measure?

Scale's SEAL leaderboard for Arabic: expert raters judge how well models answer about 1,000 prompts written by native Arabic speakers. BenchLM shows it as display-only reference data, excluded from rankings. Scale marked this board deprecated as of March 2025 and no longer adds models.

### Which model leads the published SEAL Arabic snapshot?

Gemini 1.5 Pro (August 27, 2024) currently leads the published SEAL Arabic snapshot with a score of 1147.

### How many models are evaluated on SEAL Arabic?

The November 12, 2026 snapshot contains 19 AI models.

### Does SEAL Arabic affect BenchLM's overall score?

Not directly. SEAL Arabic is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.