MATH-500

Reasoning

Curated 500-question subset of MATH for faster standardized evaluation

Higher is better

Rank	Model	Provider	Score
1	GPT-5 (high)	OpenAI	99.4%
2	Grok 3 mini Reasoning (high)	SpaceXAI	99.2%
3	OpenAI: o3	OpenAI	99.2%
4	GPT-5 (medium)	OpenAI	99.1%
5	Claude 4 Sonnet (Reasoning)	Anthropic	99.1%
6	Grok 4	SpaceXAI	99.0%
7	OpenAI: o4 Mini High	OpenAI	98.9%
8	GPT-5 (low)	OpenAI	98.7%
9	Gemini 2.5 Pro Preview (May' 25)	Google	98.6%
11	Qwen3 235B A22B 2507 (Reasoning)	Alibaba	98.4%
12	Llama Nemotron Super 49B v1.5 (Reasoning)	NVIDIA	98.3%
13	DeepSeek R1 0528 (May '25)	DeepSeek	98.3%
14	Claude 4 Opus (Reasoning)	Anthropic	98.2%
15	Gemini 2.5 Flash (Reasoning)	Google	98.1%
16	Gemini 2.5 Flash Preview (Reasoning)	Google	98.1%
17	MiniMax M1 80k	MiniMax	98.0%
18	Qwen3 235B A22B 2507 Instruct	Alibaba	98.0%
19	Gemini 2.5 Pro Preview (Mar' 25)	Google	98.0%
20	GLM-4.5 (Reasoning)	Z.ai	97.9%
21	EXAONE 4.0 32B (Reasoning)	LG AI Research	97.7%
22	Qwen3 30B A3B 2507 (Reasoning)	Alibaba	97.6%
23	Qwen3 30B A3B 2507 Instruct	Alibaba	97.5%
24	OpenAI: o3 Mini	OpenAI	97.3%
10	↳OpenAI: o3 Mini High	OpenAI	98.5%
25	MiniMax M1 40k	MiniMax	97.2%
26	MoonshotAI: Kimi K2 0711	Moonshot AI	97.1%
27	OpenAI: o1	OpenAI	97.0%
28	Gemini 2.5 Flash-Lite (Reasoning)	Google	96.9%
29	Google: Gemini 2.5 ProRetiring	Google	96.7%
30	Solar Pro 2 (Reasoning)	Upstage	96.7%
31	DeepSeek R1 (Jan '25)	DeepSeek	96.6%
32	Z.ai: GLM 4.5 Air	Z.ai	96.5%
33	Magistral Small 1	Mistral AI	96.3%
34	Qwen3 14B (Reasoning)	Alibaba	96.1%
35	Qwen3 32B (Reasoning)	Alibaba	96.1%
36	Qwen3 30B A3B (Reasoning)	Alibaba	95.9%
37	Llama 3.3 Nemotron Super 49B v1 (Reasoning)	NVIDIA	95.9%
38	Perplexity: Sonar Reasoning Pro	Perplexity	95.7%
39	Qwen/QwQ-32B	Alibaba	95.7%
40	R1 1776	Perplexity	95.4%
41	Llama 3.1 Nemotron Ultra 253B v1 (Reasoning)	NVIDIA	95.2%
42	deepseek-ai/DeepSeek-R1-Distill-Qwen-14B	DeepSeek	94.9%
43	Llama 3.1 Nemotron Nano 4B v1.1 (Reasoning)	NVIDIA	94.7%
44	Claude 3.7 Sonnet (Reasoning)	Anthropic	94.7%
45	Gemini 2.0 Flash Thinking Experimental (Jan '25)	Google	94.4%
46	o1-mini	OpenAI	94.4%
47	qwen3-coder-480b-a35b-instruct	Alibaba	94.2%
48	deepseek-v3-0324	DeepSeek	94.2%
49	Claude 4 Opus (Non-reasoning)	Anthropic	94.1%
50	deepseek-ai/DeepSeek-R1-Distill-Qwen-32B	DeepSeek	94.1%

Access this data

JSON GET /api/v1

[
  {
    "rank": 1,
    "model": "GPT-5 (high)",
    "provider": "OpenAI",
    "score": "99.4"
  },
  {
    "rank": 2,
    "model": "Grok 3 mini Reasoning (high)",
    "provider": "SpaceXAI",
    "score": "99.2"
  },
  {
    "rank": 3,
    "model": "OpenAI: o3",
    "provider": "OpenAI",
    "score": "99.2"
  }
]

Curl

$ curl -s https://rightrank.ai/api/v1/rankings/math-500.json | jq '.[:3]'

CLI brew install rightrank

$ rightrank top math-500 --limit 5
$ rightrank compare claude-opus-4.6 gpt-5.4-pro --category math-500
$ rightrank value math-500  # best performance per dollar

Claude Code Skill Skill

# In Claude Code, use the RightRank skill:
/rightrank best model for math-500
/rightrank compare top 5 by value
/rightrank cheapest model scoring above 80 on math-500