Global PIQA - Parallel (Strict Exact Match)
General QA · 2025-10-28
Global PIQA's parallel split presents aligned physical-commonsense questions with four answer options. This definition scores generated answers using exact match with strict answer extraction.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemma 4 31B IT | 82.4 |
| Qwen3.5-27B | 71.6 |
| gpt-oss-20b | 65.6 |
| Gemma 3 27B IT | 65.4 |
| Command A | 65.0 |
| Qwen3 14B | 64.1 |
| Qwen2.5 72B Instruct | 60.8 |
| Qwen3 32B | 60.8 |
| Gemma 3 12B IT | 59.2 |
| Qwen3.5-9B | 58.7 |
| Qwen3 8B | 58.4 |
| DeepSeek-R1-Distill-Qwen-32B | 56.5 |
| Qwen2.5 Instruct 32B | 56.4 |
| Gemma 2 27B IT | 54.5 |
| Qwen3 4B | 53.0 |
| DeepSeek R1 Distill Qwen 14B | 51.8 |
| Llama 3.1 70B Instruct | 51.1 |
| Qwen2.5 14B Instruct | 49.0 |
| gpt-oss-120b | 48.0 |
| Gemma 2 9B IT | 47.4 |
| Apertus 70B Instruct | 47.1 |
| Phi-4 | 46.5 |
| Aya Expanse 32B | 45.9 |
| Gemma 3 4B IT | 42.8 |
| Qwen2.5 7B Instruct | 42.1 |
| Qwen3 1.7B | 41.5 |
| C4AI Command R 08-2024 | 40.7 |
| Apertus 8B Instruct | 38.9 |
| C4AI Command R+ 08-2024 | 38.1 |
| Llama 3.1 8B Instruct | 35.8 |
| Phi-3 Medium 4K Instruct | 35.6 |
| Qwen2.5 3B Instruct | 35.4 |
| Gemma 2 2B IT | 34.1 |
| Tiny Aya Global | 33.7 |
| Phi-4-mini-instruct | 33.5 |
| Command R7B (Dec 2024) | 33.0 |
| Phi-3.5 Mini Instruct | 32.0 |
| Qwen3 0.6B | 30.3 |
| Phi-3-mini-4k-instruct | 29.0 |
| Llama 3.2 3B Instruct | 27.4 |