What does GPQA measure?

GPQA is a knowledge benchmark in the BenchGecko catalog. 72 AI models have been tested on it. Scores range from 0.5 to 19.7 out of 100.

Which model leads on GPQA?

Meta Llama 3 8B from Meta leads GPQA with a score of 19.7. The median score across 72 tested models is 4.8.

No · the top score is 19.7 out of 100 (20%). There is still meaningful room for improvement on GPQA.

Does GPQA predict performance on other benchmarks?

Yes · GPQA scores correlate 0.97 with CMMLU across 5 shared models. Models that do well on GPQA tend to do well on CMMLU.

How often is GPQA data refreshed?

BenchGecko pulls updates daily. New model scores on GPQA appear as soon as they are published by Epoch AI or the model provider.

Benchmark · KnowledgeSettled

GPQA

Name: GPQA Benchmark
Creator: BenchGecko
License: https://creativecommons.org/licenses/by/4.0/

Updated 2025-10-17

Models tested

Top score

19.7

Meta Llama 3 8B

Median

4.8

min 0.5

Top-5 spread

σ 0.8

Settled

The Frontier

Best score over time · one chart, every benchmark

Chart type

Only 19 models have been tested on GPQA · not enough history to compute a frontier yet.

Pink dots = frontier records · 0 totalClick to open model page

Full rankings

72 models tested · sorted by score

#	Model	Score	Price
1	Meta Llama 3 8B· Meta	19.7	n/a
2	HA Qwen2.5 72B Instruct Abliterated· HuiHui AI	19.4	n/a
3	Qwen2-72B· Alibaba Qwen	19.2	n/a
4	DeepSeek R1 Distill Qwen 14B· DeepSeek	18.3	n/a
5	WizardLM-2 8x22B· Microsoft	17.6	$0.62
6	Gemma 2 27B· Google DeepMind	16.7	$0.65
7	Qwen2.5 72B Instruct· Alibaba Qwen	16.7	$0.36
8	U Yi 6B· Unknown	15.6	n/a
9	Hermes 3 70B Instruct· nousresearch	14.9	$0.70
10	Gemma 2 9B· Google DeepMind	14.8	$0.03
11	Llama 3.1 70B Instruct· Meta	14.2	$0.40
12	Qwen2.5 Coder 32B Instruct· Alibaba Qwen	13.2	$0.66
13	D Dolphin 2.9.1 Yi 1.5 34b· DPHN	12.4	n/a
14	Phi 3.5 Mini Instruct· Microsoft	12.0	n/a
15	Qwen2.5 32B Instruct· Alibaba	11.7	n/a
16	Phi 3 Mini 4k Instruct· Microsoft	11.0	n/a
17	Llama 3.3 70B Instruct· Meta	10.5	$0.10
18	Qwen2.5 14B Instruct· Alibaba	10.5	n/a
19	Magnum v4 72B· anthracite-org	10.4	$3.00
20	Llama 3.1 8B Instruct· Meta	9.5	$0.02
21	Qwen2 VL 7B Instruct· Alibaba	9.3	n/a
22	Phi 4· Microsoft	9.2	$0.07
23	GLM 4 32B · z-ai	8.8	$0.10
24	U Stable Beluga 2· Unknown	8.8	n/a
25	Llama 3-70B· Meta	8.1	n/a
26	Phi 4 Mini Instruct· Microsoft	7.9	$0.08
27	Phi 4 Mini Instruct· Microsoft	7.9	n/a
28	Qwen2.5 Coder 14B Instruct· Alibaba	7.3	n/a
29	Qwen2 7B Instruct· Alibaba	6.4	n/a
30	Llama 3.1 405B· Meta	5.9	n/a
31	Hermes 2 Pro - Llama-3 8B· nousresearch	5.7	$0.14
32	Mistral 7B V0.1· Mistral AI	5.6	n/a
33	Qwen2.5 Coder 7B Instruct· Alibaba Qwen	5.6	$0.03
34	Qwen2.5 Coder 7B Instruct· Alibaba	5.6	n/a
35	Qwen2.5 7B Instruct· Alibaba Qwen	5.5	$0.10
36	Gemma 2B· Google DeepMind	4.9	n/a
37	DeepSeek R1 Distill Qwen 32B· DeepSeek	4.6	n/a
38	R1 Distill Qwen 32B· DeepSeek	4.6	$0.29
39	DeepSeek R1 Distill Qwen 7B· DeepSeek	3.9	n/a
40	Llama 3.2 3B Instruct· Meta	3.8	$0.05
41	LLaMA-13B· Meta	3.5	n/a
42	Mistral 7B Instruct V0.2· Mistral AI	3.5	n/a
43	Gemma 2 2b It· Google DeepMind	3.2	n/a
44	U StarCoder 2 15B· Unknown	3.1	n/a
45	Qwen2.5 3B Instruct· Alibaba	3.0	n/a
46	Phi 2· Microsoft	2.9	n/a
47	Falcon-180B· TII	2.8	n/a
48	Llama 3.2 1B Instruct· Meta	2.4	$0.03
49	Llama 3.2 3B Instruct (free)· Meta	2.4	$0.00
50	Phi-1.5· Microsoft	2.4	n/a
51	Llama 2 7b Hf· Meta	2.2	n/a
52	Llama 3 8B Instruct· Meta	2.1	$0.14
53	DeepSeek R1 Distill Llama 8B· DeepSeek	2.0	n/a
54	R1 Distill Llama 70B· DeepSeek	2.0	$0.80
55	Gemma 2 2b· Google DeepMind	1.7	n/a
56	Gpt2 Medium· OpenAI	1.7	n/a
57	Qwen2 1.5B Instruct· Alibaba	1.6	n/a
58	Qwen2 0.5B· Alibaba	1.4	n/a
59	Gpt2· OpenAI	1.3	n/a
60	U MPT-30B· Unknown	1.3	n/a
61	QwQ 32B· Alibaba Qwen	1.3	$0.15
62	D Distilgpt2· DistilBERT	1.2	n/a
63	Gpt2 Large· OpenAI	1.2	n/a
64	Meta Llama 3 8B Instruct· Meta	1.2	n/a
65	Qwen2.5 0.5B Instruct· Alibaba	1.2	n/a
66	Pythia 160m· eleutherai	1.1	n/a
67	L Vicuna 7b V1.5· LMSYS	1.1	n/a
68	DeepSeek R1 Distill Qwen 1.5B· DeepSeek	0.8	n/a
69	Qwen2.5 1.5B Instruct· Alibaba	0.8	n/a
70	U RedPajama-INCITE-7B-Base· Unknown	0.7	n/a
71	Gpt Neo 125m· eleutherai	0.5	n/a
72	Llama 2 7b Chat Hf· Meta	0.5	n/a