QIMMA - GAT: leaderboard
Questions from the Saudi General Aptitude Test (verbal and quantitative reasoning), as run in the QIMMA Arabic leaderboard. Scored with LightEval log-likelihood over the answer options with character-length normalization rather than generated answers, so reasoning-tuned models can land near chance; accuracy 0-100, higher is better.
Metric: Accuracy (%, log-likelihood multiple choice). Source: huggingface.co. Saturation forecast: Not forecast. 83 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | AIC-1 | 62.11 | #235 |
| 2 | Gemma 4 31B | 60.59 | #212 |
| 3 | Qwen 3.5 35B A3B Base | 60.47 | #319 |
| 4 | Qwen 2.5 32B Instruct | 56.94 | #509 |
| 5 | gemma-3-27B-pt | 56.45 | #505 |
| 6 | AceGPT-v2-32B-Chat | 56.04 | #518 |
| 7 | Qwen 2.5 72B Instruct | 55.9 | #469 |
| 8 | Qwen 3.5 397B A17B | 55.89 | #154 |
| 9 | Qwen3-8B-Base | 55.55 | #626 |
| 10 | Gemma 3 27B (IT) | 55.05 | #528 |
| 11 | Gemma 4 26B A4B | 52.04 | #360 |
| 12 | Jais-2-70B-Chat | 51.67 | #430 |
| 13 | AceGPT-v2-32B | 51.4 | #771 |
| 14 | Llama 3.3 70B Instruct | 51.13 | #541 |
| 15 | Qwen 3.5 122B A10B | 50.9 | #184 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=qimma-gat · How It Works · Data refreshed daily, snapshot 2026-11-05.