MATH Level 5 (Epoch AI) · Open Analysis

#	Model	Effort	Score	Cost / task	Harness
1	GPT-5	high	98.1%	–	Epoch AI
2	GPT-5 Mini	high	97.8%	–	Epoch AI
3	o4 mini	high	97.8%	–	Epoch AI
4	o3	high	97.8%	–	Epoch AI
5	Sonnet 4.5	high	97.7%	–	Epoch AI
6	Qwen 3 Max	max	97.1%	–	Epoch AI
7	DeepSeek R1	xhigh	96.6%	–	Epoch AI
8	o3 mini	high	96.5%	–	Epoch AI
9	Haiku 4.5	xhigh	96.4%	–	Epoch AI
10	Gemini 2.5 Pro	high	95.9%	–	Epoch AI
11	GPT-5 Nano	medium	95.2%	–	Epoch AI
12	o1	high	94.7%	–	Epoch AI
13	Grok 3 Mini	low	90.9%	–	Epoch AI
14	o1 mini	high	89.2%	–	Epoch AI
15	Grok 3	none	88.7%	–	Epoch AI
16	GPT-4.1 Mini	high	87.3%	–	Epoch AI
17	Opus 4	medium	85.0%	–	Epoch AI
18	Sonnet 4	high	84.4%	–	Epoch AI
19	Gemini 2.0 Pro	none	83.5%	–	Epoch AI
20	GPT-4.1	high	83.0%	–	Epoch AI
21	Gemini 2.0 Flash	medium	82.2%	–	Epoch AI
22	Mistral Medium 3	none	81.6%	–	Epoch AI
23	GPT-4.5	none	78.6%	–	Epoch AI
24	DeepSeek V3	none	75.5%	–	Epoch AI
25	Gemma 3 27B	none	74.0%	–	Epoch AI
26	Llama 4 Maverick	medium	73.0%	–	Epoch AI
27	Gemini 1.5 Pro	none	70.4%	–	Epoch AI
28	GPT-4.1 Nano	high	70.0%	–	Epoch AI
29	Qwen 3	high	68.9%	–	Epoch AI
30	Sonnet 3.7	medium	68.2%	–	Epoch AI
31	Qwen 2.5 Max	none	67.2%	–	Epoch AI
32	Grok 2	none	63.5%	–	Epoch AI
33	Qwen 2.5 72B	none	63.2%	–	Epoch AI
34	Llama 4 Scout	medium	62.3%	–	Epoch AI
35	Qwen 2.5 32B	none	56.1%	–	Epoch AI
36	GPT-4o mini	high	52.6%	–	Epoch AI
37	Mistral Large 2	none	50.3%	–	Epoch AI
38	GPT-4o	high	49.8%	–	Epoch AI
39	Llama 3.1 405b	none	49.8%	–	Epoch AI
40	Mistral Small 3.1	none	46.8%	–	Epoch AI
41	GPT-4	high	46.7%	–	Epoch AI
42	Mistral Small 3	none	44.8%	–	Epoch AI
43	Llama 3.3 70b	none	41.6%	–	Epoch AI
44	Llama 3.2 90b	none	39.4%	–	Epoch AI
45	Qwen 2 72B	none	39.1%	–	Epoch AI
46	Opus 3	none	37.5%	–	Epoch AI
47	Llama 3.1 70b	none	36.7%	–	Epoch AI
48	Gemma 2 27B	–	27.9%	–	Epoch AI
49	Gemini 1.5 Flash	none	25.1%	–	Epoch AI
50	Mistral Large	none	24.5%	–	Epoch AI
51	Llama 3.1 8b	medium	22.9%	–	Epoch AI
52	Llama 3 70b	none	22.7%	–	Epoch AI
53	Gemma 2 9B	–	21.0%	–	Epoch AI
54	Sonnet 3	none	18.2%	–	Epoch AI
55	Haiku 3	medium	14.9%	–	Epoch AI
56	GPT-3.5	high	11.6%	–	Epoch AI
57	Gemini 1.0 Pro	none	11.2%	–	Epoch AI
58	Mixtral 8x7B Instruct	none	9.3%	–	Epoch AI
59	Llama 3 8b	none	6.1%	–	Epoch AI
60	Mistral	none	3.6%	–	Epoch AI
61	Llama 2 70b	none	3.3%	–	Epoch AI