Last updated: just now

Arena 基准测试

LMSYS Chatbot Arena Elo Rating 排行榜

Human preference rating from 6M+ crowdsourced blind head-to-head comparisons. Users chat with two anonymous models and pick the better response.

为什么重要： The most trusted 'vibes-based' benchmark — reflects real human preferences, not just academic metrics. Widely considered the most meaningful overall ranking.

顶级模型

1,503

Claude Opus 4.6

平均评分

1,352

共102个模型

已测试模型

102

指标: Elo rating

人类基准

—

评分范围: 900–1600

模型排名

All models with a reported Arena Elo score, ranked by highest Elo rating.

排名模型评分性能

Claude Opus 4.6Anthropic

1,503

86%

1,503

Grok 4.20 BetaxAI

1,496

85%

1,496

Gemini 3.1 Pro PreviewGoogle

1,492

85%

1,492

Gemini 3 ProGoogle

1,486

84%

1,486

GPT-5.4OpenAI

1,485

84%

1,485

GPT-5.2OpenAI

1,481

83%

1,481

GPT-5.2 ChatOpenAI

1,481

83%

1,481

Gemini 3 FlashGoogle

1,474

82%

1,474

Grok 4.1 FastxAI

1,473

82%

1,473

#10

GPT-5OpenAI

1,465

81%

1,465

#11

Grok 4xAI

1,462

80%

1,462

#12

Claude Sonnet 4.6Anthropic

1,460

80%

1,460

#13

GPT-5.1OpenAI

1,456

79%

1,456

#14

Claude Sonnet 4.5Anthropic

1,452

79%

1,452

#15

Qwen3.5 397B A17BAlibaba

1,450

79%

1,450

#16

Claude Opus 4.1Anthropic

1,449

78%

1,449

#17

Gemini 2.5 ProGoogle

1,444

78%

1,444

#18

Gemini 3.1 Flash Lite PreviewGoogle

1,437

77%

1,437

#19

Claude Opus 4.5Anthropic

1,430

76%

1,430

#20

GPT-5 ChatOpenAI

1,427

75%

1,427

#21

DeepSeek V3.2 ExpDeepSeek

1,424

75%

1,424

#22

DeepSeek V3.2DeepSeek

1,423

75%

1,423

#23

Grok 4 FastxAI

1,422

75%

1,422

#24

Claude Opus 4Anthropic

1,420

74%

1,420

#25

Qwen3.5-122B-A10BAlibaba

1,419

74%

1,419

#25

DeepSeek V3.1DeepSeek

1,419

74%

1,419

#27

DeepSeek V3.1 TerminusDeepSeek

1,417

74%

1,417

#28

Qwen3 VL 235B A22B InstructAlibaba

1,416

74%

1,416

#29

o3OpenAI

1,415

74%

1,415

#30

Qwen3.5-27BAlibaba

1,410

73%

1,410

#31

MiniMax M2.5MiniMax

1,404

72%

1,404

#32

Qwen3 Next 80B A3B InstructAlibaba

1,402

72%

1,402

#33

LongCat Flash ChatMeituan

1,401

72%

1,401

#34

Qwen3.5-FlashAlibaba

1,400

71%

1,400

#35

Qwen3.5-35B-A3BAlibaba

1,398

71%

1,398

#36

Qwen3 VL 235B A22B ThinkingAlibaba

1,396

71%

1,396

#37

Gemini 2.5 FlashGoogle

1,395

71%

1,395

#38

GPT-5 MiniOpenAI

1,391

70%

1,391

#39

Step 3.5 FlashStepFun

1,389

70%

1,389

#40

Claude 3.7 Sonnet (thinking)Anthropic

1,388

70%

1,388

#41

Claude Sonnet 4Anthropic

1,387

70%

1,387

#42

o4-miniOpenAI

1,380

69%

1,380

#42

DeepSeek R1-0528DeepSeek

1,380

69%

1,380

#44

o3-miniOpenAI

1,371

67%

1,371

#45

DeepSeek R1DeepSeek

1,369

67%

1,369

#45

Qwen3 Next 80B A3B ThinkingAlibaba

1,369

67%

1,369

#47

MiniMax M1MiniMax

1,367

67%

1,367

#48

o3 Mini HighOpenAI

1,364

66%

1,364

#49

Grok 3 Mini BetaxAI

1,358

65%

1,358

#50

gpt-oss-120bOpenAI

1,355

65%

1,355

#51

Claude 3.7 SonnetAnthropic

1,354

65%

1,354

#51

Command ACohere

1,354

65%

1,354

#53

Gemini 2.0 FlashGoogle

1,352

65%

1,352

#54

o1OpenAI

1,350

64%

1,350

#55

MiniMax M2MiniMax

1,348

64%

1,348

#55

Qwen3 8BAlibaba

1,348

64%

1,348

#57

GPT-4o (2024-05-13)OpenAI

1,346

64%

1,346

#58

DeepSeek V3 (March 2025)DeepSeek

1,345

64%

1,345

#59

Grok 3xAI

1,342

63%

1,342

#59

Llama 3.3 Nemotron Super 49B V1.5NVIDIA

1,342

63%

1,342

#61

GPT-5 NanoOpenAI

1,338

63%

1,338

#61

Nova 2 LiteAmazon

1,338

63%

1,338

#63

QwQ 32BAlibaba

1,336

62%

1,336

#64

GPT-4o (2024-08-06)OpenAI

1,335

62%

1,335

#65

Olmo 3.1 32B InstructAllen AI

1,331

62%

1,331

#66

Llama 4 MaverickMeta

1,325

61%

1,325

#67

DeepSeek V3DeepSeek

1,318

60%

1,318

#67

GPT-4o-mini (2024-07-18)OpenAI

1,318

60%

1,318

#67

gpt-oss-20bOpenAI

1,318

60%

1,318

#70

Mistral Large 2407Mistral AI

1,314

59%

1,314

#71

MercuryInception

1,309

58%

1,309

#72

Olmo 3 32B ThinkAllen AI

1,307

58%

1,307

#73

o1-miniOpenAI

1,304

58%

1,304

#74

Qwen2.5 72B InstructAlibaba

1,303

58%

1,303

#75

GPT-4.1OpenAI

1,300

57%

1,300

#76

Llama 3.1 Nemotron 70B InstructNVIDIA

1,299

57%

1,299

#77

GPT-4oOpenAI

1,286

55%

1,286

#78

Olmo 3.1 32B ThinkAllen AI

1,285

55%

1,285

#79

Gemini 1.5 ProGoogle

1,281

54%

1,281

#80

Mistral Large 2Mistral AI

1,280

54%

1,280

#81

Llama 3 70B InstructMeta

1,276

54%

1,276

#82

Claude 3.5 SonnetAnthropic

1,271

53%

1,271

#82

Qwen2.5 Coder 32B InstructAlibaba

1,271

53%

1,271

#84

Grok 2xAI

1,262

52%

1,262

#84

Command R+Cohere

1,262

52%

1,262

#86

Qwen 2.5 72BAlibaba

1,261

52%

1,261

#86

Claude 3 HaikuAnthropic

1,261

52%

1,261

#88

Phi 4Microsoft

1,256

51%

1,256

#89

GPT-4 TurboOpenAI

1,255

51%

1,255

#90

Command R (08-2024)Cohere

1,250

50%

1,250

#91

Llama 3.3 70BMeta

1,243

49%

1,243

#92

Claude Haiku 4.5Anthropic

1,240

49%

1,240

#93

Claude 3 OpusAnthropic

1,232

47%

1,232

#94

Llama 3.1 405BMeta

1,229

47%

1,229

#95

Llama 3 8B InstructMeta

1,223

46%

1,223

#96

GPT-4o miniOpenAI

1,222

46%

1,222

#97

Llama 3.1 8B InstructMeta

1,212

45%

1,212

#98

Llama 3.1 70BMeta

1,198

43%

1,198

#99

Claude 3.5 HaikuAnthropic

1,178

40%

1,178

#100

Llama 3.2 3B InstructMeta

1,167

38%

1,167

#101

Mixtral 8x22BMistral AI

1,146

35%

1,146

#102

Llama 3.2 1B InstructMeta

1,111

30%

1,111

关于 Arena Elo

全名: LMSYS Chatbot Arena Elo Rating
类别: Arena
指标: Elo rating
评分范围: 900–1600
人类基准: 尚未确定
状态: 启用

Frequently Asked Questions

Arena Elo is a standardized evaluation that measures AI model performance on specific tasks. It provides comparable scores across different models, helping developers choose the right model for their needs.

Claude Opus 4.6 currently holds the top score on the Arena Elo benchmark. See our full rankings table above for the complete leaderboard with 102 models.

We update benchmark data from multiple sources including HuggingFace Open LLM Leaderboard and LMArena. Scores are refreshed regularly as new evaluations are published and new models are released.

No. While Arena Elo is an important indicator, real-world performance depends on many factors including pricing, latency, context window, and specific task requirements. We recommend using our composite score which weighs multiple benchmarks and practical factors.

LMSYS Chatbot Arena Elo Rating 排行榜

模型排名

关于 Arena Elo

相关基准测试

LMSYS Chatbot Arena Elo Rating 排行榜

模型排名

关于 Arena Elo

相关基准测试