QIMMA - FannOrFlop: leaderboard
FannOrFlop Arabic poetry understanding (classical and modern verse, literary devices and cultural context), scored in the QIMMA Arabic leaderboard with an adapted FannOrFlop evaluation as F1 (0-100, higher is better). Base checkpoints often score near zero.
Metric: F1 (%). Source: huggingface.co. Saturation forecast: Not forecast. 83 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 122B A10B | 60.54 | #184 |
| 2 | Qwen 2.5 32B Instruct | 60.1 | #509 |
| 3 | Qwen 3 235B A22B 2507 Instruct | 59.88 | #315 |
| 4 | Qwen 3.8 27B | 59.84 | #123 |
| 5 | Gemma 4 31B (IT) | 59.77 | #178 |
| 6 | Gemma 3 27B (IT) | 59.74 | #528 |
| 7 | Yehia-7B-preview | 59.64 | #659 |
| 8 | Qwen 3.5 9B | 59.57 | #391 |
| 9 | GPT-OSS-120B | 58.82 | #346 |
| 10 | c4ai-command-r7B-arabic-02-2025 | 58.55 | #727 |
| 11 | Gemma 3 12B (IT) | 57.53 | #672 |
| 12 | Qwen 2.5 72B Instruct | 57.51 | #469 |
| 13 | Qwen 3 8B | 57.47 | #678 |
| 14 | Qwen 3.6 27B | 57.21 | #182 |
| 15 | emirati-14B-v2 | 56.93 | #734 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=qimma-fannorflop · How It Works · Data refreshed daily, snapshot 2026-11-05.