QIMMA - AraDiCE-Culture: leaderboard
AraDiCE cultural-awareness multiple-choice questions about Arab regions, 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: Around February 2027. 83 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 397B A17B | 82.78 | #154 |
| 2 | Yehia-7B-preview | 81.67 | #659 |
| 3 | ALLaM-7B-Instruct-preview | 78.89 | #724 |
| 4 | Jais-2-70B-Chat | 78.89 | #430 |
| 5 | Llama 3.3 70B Instruct | 77.22 | #541 |
| 6 | Qwen 2.5 72B Instruct | 77.22 | #469 |
| 7 | Gemma 4 31B | 77.22 | #212 |
| 8 | AceGPT-v2-32B-Chat | 76.67 | #518 |
| 9 | jais-adapted-70B-chat | 76.67 | #639 |
| 10 | jais-adapted-13B-chat | 75.56 | #948 |
| 11 | Qwen 3.5 122B A10B | 74.44 | #184 |
| 12 | Fanar-1-9B-Instruct | 74.44 | #635 |
| 13 | Fanar-1-9B | 74.44 | #784 |
| 14 | AIC-1 | 73.33 | #235 |
| 15 | Gemma 4 26B A4B | 71.67 | #360 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=qimma-aradice-culture · How It Works · Data refreshed daily, snapshot 2026-11-05.