Reasoning benchmark

PIQA leaderboard

As of October 2026, GPT-4o mini has the highest published PIQA score on Noometry at 88.7%, out of 27 models with results.

Last verified October 10, 2026

About PIQA

Physical commonsense: choose the better way to accomplish a goal.

Sponsored placements are available on pages like this one. Advertise on Noometry

PIQA results by model
#	Model	Provider	Score	Source
1	GPT-4o mini	OpenAI	88.7%	Epoch AI
2	Phi-3.5-MoE	Microsoft	88.6%	Epoch AI
3	Gemini 1.5 Flash (May 2024)	Google	87.5%	Epoch AI
4	Llama 3.1-405B	Meta	85.9%	Epoch AI
5	Falcon-180B	Technology Innovation Institute	84.9%	Epoch AI
6	DeepSeek-V3	DeepSeek	84.7%	Epoch AI
7	DeepSeek-V2 (MoE-236B, May 2024)	DeepSeek	83.9%	Epoch AI
8	Gemma 2 9B	Google	83.7%	Epoch AI
9	Mixtral 8x7B	Mistral AI	83.6%	Epoch AI
10	Mistral Nemo	Mistral AI	83.5%	Epoch AI
11	Falcon-40B	Technology Innovation Institute	83%	Epoch AI
12	Mistral 7B	Mistral AI	83%	Epoch AI
13	Llama 2-70B	Meta	82.8%	Epoch AI
14	Qwen2.5 72B Instruct	Alibaba (Qwen)	82.6%	Epoch AI
15	Nemotron-4 15B	NVIDIA	82.4%	Epoch AI
16	Llama 2-34B	Meta	81.9%	Epoch AI
17	Gemma 7B	Google	81.2%	Epoch AI
18	Llama 3.1-8B	Meta	81.2%	Epoch AI
19	Phi-3.5-mini	Microsoft	81%	Epoch AI
20	Llama 2-13B	Meta	80.8%	Epoch AI
21	Falcon-7B	Technology Innovation Institute	80.3%	Epoch AI
22	Llama 13b	Meta	80.1%	Epoch AI
23	Qwen-14B	Alibaba (Qwen)	79.9%	Epoch AI
24	Llama 2-7B	Meta	78.8%	Epoch AI
25	Qwen-7B	Alibaba (Qwen)	77.9%	Epoch AI
26	Gemma 2B	Google	77.3%	Epoch AI
27	Dolly 2.0-12b	Databricks	75.4%	Epoch AI

Physical commonsense: choose the better way to accomplish a goal.

As of October 2026, GPT-4o mini has the highest published PIQA score on Noometry at 88.7%, out of 27 models with results.

Phi-3.5-MoE has the highest PIQA accuracy among open-weight models at 88.6%, ranking 2 of 27 overall.