MathArena · Open Analysis

#	Model	Effort	Score	Cost / task	Harness
1	GPT-6.1 Sol	max	90.1%	–	–
2	GPT-6 Astra	max	88.0%	–	–
3	GPT-6 Sol	max	84.9%	–	–
4	Opus 5.5	max	83.3%	–	–
5	Fable 5.1	max	82.6%	–	–
6	Opus 5	max	73.7%	–	–
7	GPT-5.6 Sol	max	66.9%	–	–
8	GPT-5.5	xhigh	63.9%	–	–
9	Qwen 3.8 Max	max	56.1%	–	–
10	GPT-5.4 Pro	xhigh	55.6%	–	–
11	Fable 5	max	55.4%	–	–
12	Kimi K3	max	50.5%	–	–
13	Grok 4.7	xhigh	46.7%	–	–
14	Muse Spark 1.3	max	46.6%	–	–
15	Muse Spark 1.1	max	46.5%	–	–
16	Opus 4.8	max	43.5%	–	–
17	DeepSeek V4.1 Flash	max	43.3%	–	–
18	GPT-5.4	xhigh	40.8%	–	–
19	Grok 4.5	max	38.5%	–	–
20	GPT-5.2	xhigh	37.4%	–	–
21	Gemini 3.8 Flash	max	35.3%	–	–
22	Gemini 3.1 Pro	high	32.6%	–	–
23	DeepSeek V4 Pro	max	29.7%	–	–
24	DeepSeek V4 Flash	max	29.6%	–	–
25	Gemini 3.7 Flash	max	29.3%	–	–
26	Kimi K2.6	max	27.8%	–	–
27	Gemini 3.5 Flash	xhigh	27.0%	–	–
28	Gemini 3.6 Flash	xhigh	26.8%	–	–
29	Opus 4.6	max	26.7%	–	–
30	GLM 5.2	max	26.4%	–	–
31	GLM 5.1	max	23.4%	–	–
32	Gemini 3 Pro	high	22.8%	–	–
33	Opus 4.7	max	22.7%	–	–
34	Gemini 3 Flash	high	22.7%	–	–
35	GLM 5	high	22.0%	–	–
36	DeepSeek V3.2	high	21.5%	–	–
37	Qwen 3.5 397B-A17B	max	21.0%	–	–
38	Kimi K2.5	max	19.6%	–	–
39	GPT-5	high	18.7%	–	–
40	GPT-5.1	high	18.7%	–	–
41	GLM 4.6	none	17.8%	–	–
42	Kimi K2	medium	17.2%	–	–
43	Qwen 3.5 27B	high	17.1%	–	–
44	Grok 4	high	16.4%	–	–
45	Grok 4.1 Fast	high	16.4%	–	–
46	gpt-oss-120b	high	16.3%	–	–
47	GPT-5 Mini	high	16.0%	–	–
48	o4 mini	high	16.0%	–	–
49	Qwen 3.5 35B-A3B	low	15.9%	–	–
50	Grok 4 Fast	none	15.8%	–	–
51	DeepSeek V3.1	none	15.3%	–	–
52	o3	high	15.1%	–	–
53	Qwen 3.6 35B	–	14.8%	–	–
54	GLM 4.5	xhigh	14.2%	–	–
55	Qwen 3	high	14.1%	–	–
56	Qwen 3.5 9B	low	13.9%	–	–
57	DeepSeek R1.0528	–	13.8%	–	–
58	gpt-oss-20b	high	13.7%	–	–
59	GPT-5 Nano	high	13.7%	–	–
60	Gemini 2.5 Pro	high	13.6%	–	–
61	Sonnet 4.5	high	13.5%	–	–
62	Qwen 3.5 4B	low	12.8%	–	–
63	GLM 4.5 Air	–	12.7%	–	–
64	Grok 3 Mini	high	12.4%	–	–
65	o3 mini	high	12.3%	–	–
66	Qwen 3 30B-A3B	none	12.1%	–	–
67	Qwen 3 235B-A22B	xhigh	11.6%	–	–
68	GLM 4.5 V	none	11.2%	–	–
69	Gemini 2.5 Flash	medium	11.1%	–	–
70	Opus 4	medium	10.7%	–	–
71	o1	medium	10.6%	–	–
72	DeepSeek R1	xhigh	9.3%	–	–
73	Qwen 3 4B	none	8.7%	–	–
74	Gemini 2.0 Flash	medium	7.9%	–	–
75	Sonnet 3.7	medium	7.3%	–	–
76	DeepSeek V3.03	–	6.9%	–	–
77	DeepSeek V3	none	4.3%	–	–
78	Qwen 3.5 2B	low	4.3%	–	–
79	Gemini 2.0 Pro	none	4.2%	–	–
80	GPT-4o	high	3.1%	–	–
81	Sonnet 3.5	medium	2.3%	–	–