LLM Leaderboard

Ranked by LiveCodeBench · source AA

Competitive programming, contamination-resistant.

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

#1 by LiveCodeBench

Gemini 3 Pro Preview (high)

91.7%

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

#	Model	Creator	LiveCode	In $/1M	Out $/1M	tok/s	LMArena
1	Gemini 3 Pro Preview (high)	Google	91.7%	$2.00	$12.0	0	1480
2	Gemini 3 Flash Preview (Reasoning)	Google	90.8%	$0.50	$3.00	224	1466
3	DeepSeek V3.2 Speciale	DeepSeek	89.6%	$0.00	$0.00	0	1424
4	GPT-5.2 (medium)	OpenAI	89.4%	$1.75	$14.0	0	1411
5	GLM-4.7 (Reasoning)	Z AI	89.4%	$0.60	$2.20	114	1436
6	GPT-5.2 (xhigh)	OpenAI	88.9%	$1.75	$14.0	77	1411
7	gpt-oss-120b (high)	OpenAI	87.8%	$0.15	$0.60	292	1366
8	Claude Opus 4.5 (Reasoning)	Anthropic	87.1%	$5.00	$25.0	56	—
9	GPT-5.1 (high)	OpenAI	86.8%	$1.25	$10.0	106	1441
10	MiMo-V2-Flash (Reasoning)	Xiaomi	86.8%	$0.10	$0.30	0	—
11	DeepSeek V3.2 (Reasoning)	DeepSeek	86.2%	$0.28	$0.42	0	1424
12	o4-mini (high)	OpenAI	85.9%	$1.10	$4.40	180	—
13	Gemini 3 Pro Preview (low)	Google	85.7%	$2.00	$12.0	0	1480
14	Kimi K2 Thinking	Kimi	85.3%	$0.60	$2.50	130	1414
15	GPT-5.1 Codex (high)	OpenAI	84.9%	$1.25	$10.0	178	1422
16	GPT-5 (high)	OpenAI	84.6%	$1.25	$10.0	110	1405
17	GPT-5 Codex (high)	OpenAI	84.0%	$1.25	$10.0	171	—
18	GPT-5 mini (high)	OpenAI	83.8%	$0.25	$2.00	103	1374
19	GPT-5.1 Codex mini (high)	OpenAI	83.6%	$0.25	$2.00	213	1422
20	Grok 4 Fast (Reasoning)	SpaceXAI	83.2%	$0.20	$0.50	0	1398
21	MiniMax-M2	MiniMax	82.6%	$0.30	$1.20	73	1342
22	Grok 4.1 Fast (Reasoning)	SpaceXAI	82.2%	$0.00	$0.00	0	1408
23	Grok 4	SpaceXAI	81.9%	$5.50	$27.5	0	1452
24	ERNIE 5.0 Thinking Preview	Baidu	81.2%	$0.00	$0.00	0	—
25	MiniMax-M2.1	MiniMax	81.0%	$0.30	$1.20	73	1391
26	o3	OpenAI	80.8%	$2.00	$8.00	135	—
27	Apriel-v1.6-15B-Thinker	ServiceNow	80.7%	$0.00	$0.00	0	—
28	Gemini 2.5 Pro	Google	80.1%	$1.25	$10.0	145	1457
29	DeepSeek V3.1 Terminus (Reasoning)	DeepSeek	79.8%	$1.64	$2.75	0	1419
30	Gemini 3 Flash Preview (Non-reasoning)	Google	79.7%	$0.50	$3.00	216	1466
31	DeepSeek V3.2 Exp (Reasoning)	DeepSeek	78.9%	$0.28	$0.42	0	1424
32	GPT-5 nano (high)	OpenAI	78.9%	$0.05	$0.40	161	1320
33	Qwen3 235B A22B 2507 (Reasoning)	Alibaba	78.8%	$0.70	$8.40	62	1366
34	DeepSeek V3.1 (Reasoning)	DeepSeek	78.4%	$0.59	$1.69	0	1419
35	Qwen3 Next 80B A3B (Reasoning)	Alibaba	78.4%	$0.50	$6.00	177	—
36	Gemini 2.5 Pro Preview (Mar' 25)	Google	77.8%	$0.00	$0.00	0	1457
37	INTELLECT-3	Prime Intellect	77.7%	$0.00	$0.00	0	1356
38	gpt-oss-20b (high)	OpenAI	77.7%	$0.05	$0.20	218	1288
39	Gemini 2.5 Pro Preview (May' 25)	Google	77.0%	$1.25	$10.0	0	1457
40	DeepSeek R1 0528 (May '25)	DeepSeek	77.0%	$1.35	$4.20	0	1373
41	K-EXAONE (Reasoning)	LG AI Research	76.8%	$0.00	$0.00	0	—
42	Qwen3 Max	Alibaba	76.7%	$1.20	$6.00	53	1439
43	Doubao Seed Code	ByteDance Seed	76.6%	$0.00	$0.00	0	—
44	Seed-OSS-36B-Instruct	ByteDance Seed	76.5%	$0.21	$0.57	39	—
45	GPT-5 (low)	OpenAI	76.3%	$1.25	$10.0	83	—
46	GPT-5 nano (medium)	OpenAI	76.3%	$0.05	$0.40	162	—
47	Magistral Medium 1.2	Mistral	75.0%	$2.00	$5.00	43	—
48	KAT-Coder-Pro V1	KwaiKAT	74.7%	$0.00	$0.00	0	—
49	EXAONE 4.0 32B (Reasoning)	LG AI Research	74.7%	$0.00	$0.00	0	—
50	NVIDIA Nemotron 3 Nano 30B A3B (Reasoning)	NVIDIA	74.1%	$0.05	$0.20	159	—

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.