Open LLM Leaderboard - GPQA: leaderboard

Metric: Score. Source: huggingface.co. 4576 models tracked.

Top models

#ModelScore
170B-L3.3-Cirrus-x126.62
2Mistral Large 2 (Nov) Instruct (2411)24.94
3Qwen 2.5 32B21.59
4Phi-420.81
5Qwen 2.5 72B20.69
6phi-4-14B20.47
7calme-3.2-instruct-78B20.36
8Llama 3 70B19.69
9Saka-14B19.46
10Qwen 2 72B19.24
11Llama-3.1-Nemotron-lorablated-70B18.79
12Llama 3.1 70B18.34
13DeepSeek R1 Distill Qwen 14B18.34
14Qwen 2 VL 72B Instruct18.34
15Mistral-Small-24B-Base-250118.34

Interactive version: theaggregate.ai/benchmark?slug=open-llm-leaderboard-gpqa · How It Works · Data refreshed daily, snapshot 2026-09-05.