MATH 500 (Vals) · Open Analysis

#	Model	Effort	Score	Cost / task	Harness
1	Gemini 3 Pro	high	96.4%	$0.051	–
2	Grok 4	high	96.2%	$0.037	–
3	GPT-5	high	96.0%	$0.020	–
4	Opus 4.1	high	95.4%	$0.202	–
5	Gemini 2.5 Pro	high	95.2%	$0.0079	–
6	GPT-5 Mini	high	94.8%	$0.0037	–
7	gpt-oss-120b	high	94.8%	$0.0014	–
8	o3	high	94.6%	$0.011	–
9	Qwen 3 235B	–	94.6%	$0.0034	–
10	o4 mini	high	94.2%	$0.0048	–
11	Grok 3 Mini	high	94.2%	$0.0019	–
12	gpt-oss-20b	high	94.2%	$0.0010	–
13	Kimi K2	medium	94.2%	$0.0030	–
14	GLM 4.5	xhigh	94.0%	$0.013	–
15	GPT-5 Nano	high	93.8%	$0.0017	–
16	Sonnet 4	high	93.8%	$0.068	–
17	DeepSeek R1	xhigh	92.2%	$0.011	–
18	o3 mini	high	91.8%	$0.0077	–
19	Gemini 2.5 Flash	medium	91.8%	$0.011	–
20	Sonnet 3.7	medium	91.6%	$0.125	–
21	Llama 3.3	none	91.4%	–	–
22	Opus 4	medium	90.4%	$0.045	–
23	o1	high	90.4%	$0.095	–
24	Grok 3	none	89.8%	$0.013	–
25	Gemini 2.0 Flash	medium	89.0%	$0.0002	–
26	MiniMax M2.1	–	89.0%	$0.0046	–
27	DeepSeek V3	none	88.6%	$0.0007	–
28	GPT-4.1 Mini	high	88.0%	$0.0012	–
29	GPT-4.1	high	87.2%	$0.0094	–
30	Mistral Medium 3.1	none	87.0%	$0.0018	–
31	Llama 4 Maverick	medium	85.2%	$0.0006	–
32	Gemini 1.5 Pro	none	82.8%	$0.0022	–
33	GPT-4.1 Nano	high	80.2%	$0.0003	–
34	Llama 4 Scout	medium	79.2%	$0.0004	–
35	Gemini 1.5 Flash	none	78.8%	$0.0001	–
36	Grok 2	none	78.4%	$0.0063	–
37	Command A	none	76.2%	$0.0060	–
38	GPT-4o	high	75.2%	$0.0056	–
39	Mistral Large 11	–	74.4%	$0.0033	–
40	GPT-4o mini	high	72.6%	$0.0003	–
41	Claude 3.5 Sonnet Latest	–	72.4%	$0.0050	–
42	Llama 3.1	max	71.4%	$0.0030	–
43	Mistral Small 02	–	70.6%	$0.0003	–
44	Mistral Small 3.1	none	68.4%	$0.0002	–
45	Haiku 3.5	medium	64.2%	$0.0015	–
46	Jamba 1.6 Large	none	54.8%	$0.0040	–
47	Jamba 1.6 Mini	none	25.4%	$0.0003	–