MMLU-Pro

General

Harder MMLU variant with 10 answer choices and more reasoning-focused questions

Source ↗ Higher is better

Rank	Model	Provider	Score
1	Gemini 3 Pro Preview (high)	Google	89.8%
2	Gemini 3 Pro Preview (low)	Google	89.5%
3	Claude Opus 4.5 (Reasoning)	Anthropic	89.5%
4	Gemini 3 Flash Preview (Reasoning)	Google	89.0%
5	Claude Opus 4.5 (Non-reasoning)	Anthropic	88.9%
6	Gemini 3 Flash Preview (Non-reasoning)	Google	88.2%
7	Claude 4.1 Opus (Reasoning)	Anthropic	88.0%
8	MiniMax: MiniMax M2.1Historical	MiniMax	87.5%
9	Claude 4.5 Sonnet (Reasoning)	Anthropic	87.5%
10	GPT-5.2 (xhigh)	OpenAI	87.4%
11	Claude 4 Opus (Reasoning)	Anthropic	87.3%
12	GPT-5 (high)	OpenAI	87.1%
13	GPT-5.1 (high)	OpenAI	87.0%
14	GPT-5 (medium)	OpenAI	86.7%
15	Grok 4	SpaceXAI	86.6%
16	GPT-5 Codex (high)	OpenAI	86.5%
17	DeepSeek V3.2 Speciale	DeepSeek	86.3%
18	DeepSeek V3.2 (Reasoning)	DeepSeek	86.2%
19	Google: Gemini 2.5 ProRetiring	Google	86.2%
20	GPT-5 (low)	OpenAI	86.0%
21	Claude 4 Opus (Non-reasoning)	Anthropic	86.0%
22	GPT-5.1 Codex (high)	OpenAI	86.0%
23	Claude 4.5 Sonnet (Non-reasoning)	Anthropic	86.0%
24	GPT-5.2 (medium)	OpenAI	85.9%
25	Gemini 2.5 Pro Preview (Mar' 25)	Google	85.8%
26	GLM-4.7 (Reasoning)	Z.ai	85.6%
27	Doubao Seed Code	ByteDance Seed	85.4%
28	grok-4-1-fast-reasoning	SpaceXAI	85.4%
29	OpenAI: o3	OpenAI	85.3%
30	DeepSeek V3.1 (Reasoning)	DeepSeek	85.1%
31	DeepSeek V3.1 Terminus (Reasoning)	DeepSeek	85.1%
32	DeepSeek V3.2 Exp (Reasoning)	DeepSeek	85.0%
33	grok-4-fast-reasoning	SpaceXAI	85.0%
34	Cogito v2.1 (Reasoning)	Deep Cogito	84.9%
35	DeepSeek R1 0528 (May '25)	DeepSeek	84.9%
36	MoonshotAI: Kimi K2 Thinking	Moonshot AI	84.8%
37	o1-preview	OpenAI	84.8%
38	DeepSeek R1 (Jan '25)	DeepSeek	84.4%
39	MiMo-V2-Flash (Reasoning)	Xiaomi	84.3%
40	Qwen3 235B A22B 2507 (Reasoning)	Alibaba	84.3%
41	Claude 4 Sonnet (Reasoning)	Anthropic	84.2%
42	Gemini 2.5 Flash Preview (Sep '25) (Reasoning)	Google	84.2%
43	Qwen: Qwen3 MaxHistorical	Alibaba	84.1%
44	OpenAI: o1	OpenAI	84.1%
45	K-EXAONE (Reasoning)	LG AI Research	83.8%
46	qwen3-max-preview	Alibaba	83.8%
47	DeepSeek V3.2 (Non-reasoning)	DeepSeek	83.7%
48	Claude 4 Sonnet (Non-reasoning)	Anthropic	83.7%
49	GPT-5 mini (high)	OpenAI	83.7%
50	Claude 3.7 Sonnet (Reasoning)	Anthropic	83.7%

Access this data

JSON GET /api/v1

[
  {
    "rank": 1,
    "model": "Gemini 3 Pro Preview (high)",
    "provider": "Google",
    "score": "89.8"
  },
  {
    "rank": 2,
    "model": "Gemini 3 Pro Preview (low)",
    "provider": "Google",
    "score": "89.5"
  },
  {
    "rank": 3,
    "model": "Claude Opus 4.5 (Reasoning)",
    "provider": "Anthropic",
    "score": "89.5"
  }
]

Curl

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

CLI brew install rightrank

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

Claude Code Skill Skill

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