LLM Leaderboard

Ranked by MMLU-Pro · source AA

Multi-domain knowledge & reasoning. Harder than MMLU.

IntelligenceAA LMArenaLMArena CodingAA MathAA MMLU-ProAA GPQAAA LiveCodeAA MATH-500AA AIMEAA HLEAA SciCodeAA

#1 by MMLU-Pro

Gemini 3 Pro Preview (high)

89.8%

Google · in $2.00 / out $12.0 per 1M · 0 tok/s

#	Model	Creator	MMLU-Pro	In $/1M	Out $/1M	tok/s	LMArena
1	Gemini 3 Pro Preview (high)	Google	89.8%	$2.00	$12.0	0	1480
2	Claude Opus 4.5 (Reasoning)	Anthropic	89.5%	$5.00	$25.0	56	—
3	Gemini 3 Pro Preview (low)	Google	89.5%	$2.00	$12.0	0	1480
4	Gemini 3 Flash Preview (Reasoning)	Google	89.0%	$0.50	$3.00	224	1466
5	Claude Opus 4.5 (Non-reasoning)	Anthropic	88.9%	$5.00	$25.0	50	—
6	Gemini 3 Flash Preview (Non-reasoning)	Google	88.2%	$0.50	$3.00	216	1466
7	Claude 4.1 Opus (Reasoning)	Anthropic	88.0%	$15.0	$75.0	35	—
8	Claude 4.5 Sonnet (Reasoning)	Anthropic	87.5%	$3.00	$15.0	49	—
9	MiniMax-M2.1	MiniMax	87.5%	$0.30	$1.20	73	1391
10	GPT-5.2 (xhigh)	OpenAI	87.4%	$1.75	$14.0	77	1411
11	Claude 4 Opus (Reasoning)	Anthropic	87.3%	$15.0	$75.0	0	—
12	GPT-5 (high)	OpenAI	87.1%	$1.25	$10.0	110	1405
13	GPT-5.1 (high)	OpenAI	87.0%	$1.25	$10.0	106	1441
14	GPT-5 (medium)	OpenAI	86.7%	$1.25	$10.0	82	—
15	Grok 4	SpaceXAI	86.6%	$5.50	$27.5	0	1452
16	GPT-5 Codex (high)	OpenAI	86.5%	$1.25	$10.0	171	—
17	DeepSeek V3.2 Speciale	DeepSeek	86.3%	$0.00	$0.00	0	1424
18	DeepSeek V3.2 (Reasoning)	DeepSeek	86.2%	$0.28	$0.42	0	1424
19	Gemini 2.5 Pro	Google	86.2%	$1.25	$10.0	145	1457
20	GPT-5.1 Codex (high)	OpenAI	86.0%	$1.25	$10.0	178	1422
21	GPT-5 (low)	OpenAI	86.0%	$1.25	$10.0	83	—
22	Claude 4.5 Sonnet (Non-reasoning)	Anthropic	86.0%	$3.00	$15.0	42	—
23	Claude 4 Opus (Non-reasoning)	Anthropic	86.0%	$15.0	$75.0	0	—
24	GPT-5.2 (medium)	OpenAI	85.9%	$1.75	$14.0	0	1411
25	Gemini 2.5 Pro Preview (Mar' 25)	Google	85.8%	$0.00	$0.00	0	1457
26	GLM-4.7 (Reasoning)	Z AI	85.6%	$0.60	$2.20	114	1436
27	Grok 4.1 Fast (Reasoning)	SpaceXAI	85.4%	$0.00	$0.00	0	1408
28	Doubao Seed Code	ByteDance Seed	85.4%	$0.00	$0.00	0	—
29	o3	OpenAI	85.3%	$2.00	$8.00	135	—
30	DeepSeek V3.1 Terminus (Reasoning)	DeepSeek	85.1%	$1.64	$2.75	0	1419
31	DeepSeek V3.1 (Reasoning)	DeepSeek	85.1%	$0.59	$1.69	0	1419
32	Grok 4 Fast (Reasoning)	SpaceXAI	85.0%	$0.20	$0.50	0	1398
33	DeepSeek V3.2 Exp (Reasoning)	DeepSeek	85.0%	$0.28	$0.42	0	1424
34	DeepSeek R1 0528 (May '25)	DeepSeek	84.9%	$1.35	$4.20	0	1373
35	Cogito v2.1 (Reasoning)	Deep Cogito	84.9%	$1.25	$1.25	0	—
36	Kimi K2 Thinking	Kimi	84.8%	$0.60	$2.50	130	1414
37	DeepSeek R1 (Jan '25)	DeepSeek	84.4%	$1.68	$4.70	0	1373
38	MiMo-V2-Flash (Reasoning)	Xiaomi	84.3%	$0.10	$0.30	0	—
39	Qwen3 235B A22B 2507 (Reasoning)	Alibaba	84.3%	$0.70	$8.40	62	1366
40	Claude 4 Sonnet (Reasoning)	Anthropic	84.2%	$3.00	$15.0	0	—
41	Gemini 2.5 Flash Preview (Sep '25) (Reasoning)	Google	84.2%	$0.00	$0.00	0	1417
42	Qwen3 Max	Alibaba	84.1%	$1.20	$6.00	53	1439
43	o1	OpenAI	84.1%	$15.0	$60.0	126	—
44	K-EXAONE (Reasoning)	LG AI Research	83.8%	$0.00	$0.00	0	—
45	Qwen3 Max (Preview)	Alibaba	83.8%	$1.20	$6.00	57	1439
46	Claude 3.7 Sonnet (Reasoning)	Anthropic	83.7%	$0.00	$0.00	0	—
47	Claude 4 Sonnet (Non-reasoning)	Anthropic	83.7%	$3.00	$15.0	0	—
48	GPT-5 mini (high)	OpenAI	83.7%	$0.25	$2.00	103	1374
49	DeepSeek V3.2 (Non-reasoning)	DeepSeek	83.7%	$0.28	$0.42	0	1424
50	Gemini 2.5 Pro Preview (May' 25)	Google	83.7%	$1.25	$10.0	0	1457

Quality, pricing, and speed benchmarks via Artificial Analysis. Human preference Elo via LMArena. Refreshed daily at 06:00 UTC · last sync Sun, 19 Jul 2026 06:00:01 GMT.

About the metrics

Intelligence Index

Artificial Analysis composite of reasoning, knowledge, coding, math.

LMArena Elo

LMArena

Human preference Elo from chatbot arena head-to-head votes.

Coding Index

AA composite — LiveCodeBench, SciCode, etc.

Math Index

AA composite — MATH-500, AIME, etc.

MMLU-Pro

Multi-domain knowledge & reasoning. Harder than MMLU.

GPQA Diamond

Graduate-level science Q&A. Strong reasoning signal.

LiveCodeBench

Competitive programming, contamination-resistant.

MATH-500

Subset of the MATH benchmark.

AIME

American Invitational Mathematics Examination — olympiad math.

Humanity's Last Exam

Frontier reasoning benchmark; most models still score low.

SciCode

Scientific coding tasks.