Mizan: leaderboard
Mizan, a national benchmark for Iraqi Arabic and the Iraqi civic context: the frozen pilot board's auto-scored overall across the Modern Standard Arabic and Iraqi tracks (generative axes await human judging). Higher is better. Source: arxiv.org/abs/2609.13980.
Metric: Overall (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 27 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 90.9 | #159 |
| 2 | GPT-5 | 90.9 | #105 |
| 3 | GPT-6 Astra | 90.7 | #6 |
| 4 | GPT-4o | 90.2 | #357 |
| 5 | Gemini 2.5 Flash | 90.1 | #255 |
| 6 | Claude Sonnet 5 | 89.9 | #65 |
| 7 | Gemma 4 31B (IT) | 89.5 | #178 |
| 8 | Claude Opus 5 | 89.2 | #16 |
| 9 | Claude Sonnet 4.5 | 89.1 | #148 |
| 10 | GPT-OSS-120B | 88.2 | #346 |
| 11 | Mistral Large 3 | 88.2 | #410 |
| 12 | Claude Haiku 4.5 | 88.1 | #287 |
| 13 | Gemma 3 27B (IT) | 87.6 | #528 |
| 14 | Llama 4 Maverick | 86.8 | #486 |
| 15 | Claude Fable 5 | 86.4 | #20 |
No result here: #3 Claude Opus 5.5, #7 GPT-6.1 Sol, #9 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=mizan · How It Works · Data refreshed daily, snapshot 2026-11-05.