physics benchmark
PIQA
PIQA (Physical Interaction: Question Answering) is a benchmark dataset for physical commonsense reasoning in natural language. It tests AI systems' ability to answer questions requiring physical world knowledge through multiple choice questions with everyday situations, focusing on atypical solutions inspired by instructables.com. The dataset contains 21,000 multiple choice questions where models must choose the most appropriate solution for physical interactions.
Updated Aug 7, 2026
Published models11
Registry coverage11
MetricScore
EvidenceB
PIQA leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | MI | 88.6% | 100.0% | 11 | C | |
| 2 | NR | 84.4% | 90.0% | 11 | C | |
| 3 | GO | 83.2% | 80.0% | 11 | C | |
| 4 | GO | 81.7% | 70.0% | 11 | C | |
| 5 | GO | 81.0% | 60.0% | 11 | C | |
| 6 | GO | 81.0% | 50.0% | 11 | C | |
| 7 | MI | 81.0% | 40.0% | 11 | C | |
| 8 | GO | 78.9% | 30.0% | 11 | C | |
| 9 | GO | 78.9% | 20.0% | 11 | C | |
| 10 | MI | 77.6% | 10.0% | 11 | C | |
| 11 | BA | 55.2% | 0.0% | 11 | C |
Score distribution
Top published rows on the benchmark's original scale.
PIQA highlights
The top published results on this benchmark's own scale.
What is PIQA?
Definition and scoring fields from the benchmark registry.
PIQA (Physical Interaction: Question Answering) is a benchmark dataset for physical commonsense reasoning in natural language. It tests AI systems' ability to answer questions requiring physical world knowledge through multiple choice questions with everyday situations, focusing on atypical solutions inspired by instructables.com. The dataset contains 21,000 multiple choice questions where models must choose the most appropriate solution for physical interactions.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- PIQA
- Modality
- text
- Primary category
- physics
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- piqa|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about PIQA.
Which model scores highest on PIQA?
Phi-3.5-MoE-instruct is currently ranked first with 88.6%.
What does PIQA measure?
PIQA (Physical Interaction: Question Answering) is a benchmark dataset for physical commonsense reasoning in natural language. It tests AI systems' ability to answer questions requiring physical world knowledge through multiple choice questions with everyday situations, focusing on atypical solutions inspired by instructables.com. The dataset contains 21,000 multiple choice questions where models must choose the most appropriate solution for physical interactions.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
11 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is marked as eligible for the current LLMBoard capability methodology.