TurkBench - Metaphors and Idioms: leaderboard
Metric: Accuracy (%). Source: huggingface.co. 38 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.5 397B A17B | 93.33 |
| 2 | Qwen 3.5 122B A10B | 90 |
| 3 | Gemma 3 27B (IT) | 85.33 |
| 4 | Qwen 3.5 27B | 85.33 |
| 5 | Qwen 3.5 35B A3B | 85.33 |
| 6 | MiniMax-M2.5 | 83.33 |
| 7 | GLM-5 FP8 | 81.33 |
| 8 | GLM-4.6 | 78 |
| 9 | Gemma 3 12B (IT) | 74.67 |
| 10 | MiniMax-M2 | 70.67 |
| 11 | Qwen 3 30B A3B 2507 Instruct | 68.66 |
| 12 | GPT-OSS-120B | 68 |
| 13 | Gemma 2 9B (IT) | 68 |
| 14 | MiMo-V2-Flash | 68 |
| 15 | Tongyi DeepResearch 30B A3B | 68 |
Interactive version: theaggregate.ai/benchmark?slug=turkbench-metaphors-and-idioms · How It Works · Data refreshed daily, snapshot 2026-09-19.