MATH-500

Artificial Analysis · 50 models

Rank	Model	Score
1	GPT-5 (High)	0.994
2	Grok 3 mini Reasoning (High)	0.992
2	o3	0.992
4	GPT-5 (Medium)	0.9913
5	Claude 4 Sonnet (Reasoning)	0.9907
6	Grok 4	0.99
7	o4-mini (High)	0.9887
8	GPT-5 (Low)	0.9873
9	Gemini 2.5 Pro Preview (May' 25)	0.986
10	o3-mini (High)	0.9847
11	Qwen3 235B A22B 2507 (Reasoning)	0.984
12	Llama Nemotron Super 49B v1.5 (Reasoning)	0.9833
13	DeepSeek R1 0528 (May '25)	0.9827
14	Claude 4 Opus (Reasoning)	0.982
15	Gemini 2.5 Flash (Reasoning)	0.9813
16	Gemini 2.5 Flash Preview (Reasoning)	0.9807
17	Gemini 2.5 Pro Preview (Mar' 25)	0.98
17	MiniMax M1 80k	0.98
17	Qwen3 235B A22B 2507 Instruct	0.98
20	GLM-4.5 (Reasoning)	0.9787
21	EXAONE 4.0 32B (Reasoning)	0.9767
22	Qwen3 30B A3B 2507 (Reasoning)	0.976
23	Qwen3 30B A3B 2507 Instruct	0.9753
24	o3-mini	0.9733
25	MiniMax M1 40k	0.972
26	Kimi K2	0.9713
27	o1	0.97
28	Gemini 2.5 Flash-Lite (Reasoning)	0.9687
29	Gemini 2.5 Pro	0.9673
30	Solar Pro 2 (Reasoning)	0.9667
31	DeepSeek R1 (Jan '25)	0.966
32	GLM-4.5-Air	0.9653
33	Magistral Small 1	0.9627
34	Qwen3 14B (Reasoning)	0.9613
35	Qwen3 32B (Reasoning)	0.9607
36	Qwen3 30B A3B (Reasoning)	0.9593
37	Llama 3.3 Nemotron Super 49B v1 (Reasoning)	0.9587
38	QwQ 32B	0.9573
38	Sonar Reasoning Pro	0.9573
40	R1 1776	0.954
41	Llama 3.1 Nemotron Ultra 253B v1 (Reasoning)	0.952
42	DeepSeek R1 Distill Qwen 14B	0.9487
43	Claude 3.7 Sonnet (Reasoning)	0.9467
43	Llama 3.1 Nemotron Nano 4B v1.1 (Reasoning)	0.9467
45	Gemini 2.0 Flash Thinking Experimental (Jan '25)	0.944
45	o1-mini	0.944
47	DeepSeek V3 0324	0.942
47	Qwen3 Coder 480B A35B Instruct	0.942
49	Claude 4 Opus (Non-reasoning)	0.9407
49	DeepSeek R1 Distill Qwen 32B	0.9407