最近更新: 1h ago

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

1,507

Claude Fable 5

平均评分

1,377

共129个模型

已测试模型

129

指标: Elo rating

人类基准

评分范围: 900–1600

Arena Elo Scores - Top 25 Models

Ranked by Arena Elo score

LMMarketCap.com

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 Fable 5 Anthropic

1,507

87%

1,507

Claude Opus 4.6 Anthropic

1,503

86%

1,503

Qwen3.8 Max Alibaba

1,497

85%

1,497

Gemini 3.1 Pro Google

1,494

85%

1,494

Claude Opus 4.7 Anthropic

1,491

84%

1,491

Muse Spark 1.1 meta

1,487

84%

1,487

Gemini 3 Pro Google

1,486

84%

1,486

GPT-5.4 OpenAI

1,485

84%

1,485

Gemini 3.6 Flash Google

1,485

84%

1,485

#10

GPT-5.2 OpenAI

1,481

83%

1,481

#11

Claude Opus 4.8 Anthropic

1,479

83%

1,479

#12

Gemini 3.5 Flash Google

1,477

82%

1,477

#13

GPT-5.2 Chat OpenAI

1,476

82%

1,476

#14

GPT-5.1 OpenAI

1,475

82%

1,475

#14

GPT-5.5 OpenAI

1,475

82%

1,475

#16

Gemini 3 Flash Google

1,474

82%

1,474

#17

Grok 4.5 xAI

1,468

81%

1,468

#17

GLM 5.1 Zhipu AI

1,468

81%

1,468

#19

MiMo-V2.5-Pro Xiaomi

1,467

81%

1,467

#20

GPT-5 OpenAI

1,465

81%

1,465

#21

DeepSeek V4 Pro DeepSeek

1,463

80%

1,463

#22

Grok 4 xAI

1,462

80%

1,462

#23

Kimi K2.6 Moonshot AI

1,461

80%

1,461

#24

Claude Sonnet 4.6 Anthropic

1,460

80%

1,460

#24

Qwen3.6 Max Preview Alibaba

1,460

80%

1,460

#26

Gemini 3.5 Flash Lite Google

1,459

80%

1,459

#27

Qwen3.7 Plus Alibaba

1,458

80%

1,458

#28

GLM 5 Zhipu AI

1,457

80%

1,457

#29

Hy3 Tencent

1,453

79%

1,453

#30

Claude Sonnet 4.5 Anthropic

1,452

79%

1,452

#31

Gemma 4 31B Google

1,451

79%

1,451

#32

Claude Opus 4.1 Anthropic

1,449

78%

1,449

#33

Grok 4.3 xAI

1,446

78%

1,446

#34

MiniMax M3 MiniMax

1,445

78%

1,445

#35

Gemini 2.5 Pro Google

1,444

78%

1,444

#36

Qwen3.6 Plus Alibaba

1,443

78%

1,443

#37

Qwen3.5 397B A17B Alibaba

1,442

77%

1,442

#37

GLM 4.7 Zhipu AI

1,442

77%

1,442

#37

Inkling thinkingmachines

1,442

77%

1,442

#40

Gemma 4 26B A4B Google

1,438

77%

1,438

#41

MiMo-V2.5 Xiaomi

1,434

76%

1,434

#42

DeepSeek V4 Flash DeepSeek

1,433

76%

1,433

#42

GLM 5V Turbo Zhipu AI

1,433

76%

1,433

#44

Gemini 3.1 Flash Lite Preview Google

1,432

76%

1,432

#45

Inkling Small thinkingmachines

1,431

76%

1,431

#46

Claude Opus 4.5 Anthropic

1,430

76%

1,430

#47

Mistral Medium 3.5 Mistral AI

1,427

75%

1,427

#48

DeepSeek V3.2 DeepSeek

1,425

75%

1,425

#48

GLM 4.6 Zhipu AI

1,425

75%

1,425

#50

DeepSeek V3.2 Exp DeepSeek

1,423

75%

1,423

#51

Claude Opus 4 Anthropic

1,420

74%

1,420

#52

DeepSeek V3.1 DeepSeek

1,418

74%

1,418

#53

Qwen3.5-122B-A10B Alibaba

1,417

74%

1,417

#54

MiniMax M2.7 MiniMax

1,416

74%

1,416

#55

o3 OpenAI

1,415

74%

1,415

#55

Qwen3 VL 235B A22B Instruct Alibaba

1,415

74%

1,415

#55

DeepSeek V3.1 Terminus DeepSeek

1,415

74%

1,415

#58

Hy3 preview Tencent

1,412

73%

1,412

#59

GLM 4.5 Zhipu AI

1,411

73%

1,411

#60

Qwen3.5-27B Alibaba

1,408

73%

1,408

#61

Qwen3 Next 80B A3B Instruct Alibaba

1,401

72%

1,401

#62

Qwen3.5-Flash Alibaba

1,397

71%

1,397

#63

Gemini 2.5 Flash Google

1,395

71%

1,395

#63

Qwen3 VL 235B A22B Thinking Alibaba

1,395

71%

1,395

#63

Qwen3.5-35B-A3B Alibaba

1,395

71%

1,395

#66

Step 3.5 Flash StepFun

1,394

71%

1,394

#67

GPT-5 Mini OpenAI

1,390

70%

1,390

#67

MiniMax M2.5 MiniMax

1,390

70%

1,390

#69

Claude Sonnet 4 Anthropic

1,387

70%

1,387

#70

GPT-4.1 Mini OpenAI

1,383

69%

1,383

#71

o4-mini OpenAI

1,380

69%

1,380

#71

DeepSeek R1-0528 DeepSeek

1,380

69%

1,380

#73

GLM 4.6V Zhipu AI

1,377

68%

1,377

#74

GLM 4.5 Air Zhipu AI

1,373

68%

1,373

#75

o3-mini OpenAI

1,371

67%

1,371

#76

DeepSeek R1 DeepSeek

1,369

67%

1,369

#76

Qwen3 Next 80B A3B Thinking Alibaba

1,369

67%

1,369

#76

Trinity Large Thinking arcee-ai

1,369

67%

1,369

#79

GLM 4.7 Flash Zhipu AI

1,368

67%

1,368

#80

MiniMax M1 MiniMax

1,364

66%

1,364

#80

o3 Mini High OpenAI

1,364

66%

1,364

#82

Claude 3.7 Sonnet Anthropic

1,354

65%

1,354

#82

GLM 4.5V Zhipu AI

1,354

65%

1,354

#84

Gemini 2.0 Flash Google

1,352

65%

1,352

#84

gpt-oss-120b OpenAI

1,352

65%

1,352

#86

o1 OpenAI

1,350

64%

1,350

#87

Qwen3 8B Alibaba

1,347

64%

1,347

#87

Mercury 2 Inception

1,347

64%

1,347

#89

MiniMax M2 MiniMax

1,346

64%

1,346

#90

DeepSeek V3 (March 2025)DeepSeek

1,345

64%

1,345

#91

Grok 3 xAI

1,342

63%

1,342

#92

GPT-5 Nano OpenAI

1,337

62%

1,337

#92

Nova 2 Lite Amazon

1,337

62%

1,337

#94

o1 Preview OpenAI

1,334

62%

1,334

#95

Llama 4 Maverick Meta

1,325

61%

1,325

#96

GPT-4.1 Nano OpenAI

1,322

60%

1,322

#97

DeepSeek V3 DeepSeek

1,318

60%

1,318

#97

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

1,318

60%

1,318

#99

gpt-oss-20b OpenAI

1,317

60%

1,317

#100

Mistral Large 2407 Mistral AI

1,314

59%

1,314

#101

Granite 4.1 8B IBM

1,306

58%

1,306

#102

Olmo 3 32B Think Allen AI

1,305

58%

1,305

#103

o1-mini OpenAI

1,304

58%

1,304

#104

GPT-4.1 OpenAI

1,300

57%

1,300

#105

GPT-4o OpenAI

1,286

55%

1,286

#106

Gemini 1.5 Pro Google

1,281

54%

1,281

#107

Mistral Large 2 Mistral AI

1,280

54%

1,280

#108

GPT-4 OpenAI

1,275

54%

1,275

#109

Claude 3.5 Sonnet Anthropic

1,271

53%

1,271

#109

Qwen 2.5 Coder 32B Alibaba

1,271

53%

1,271

#111

Grok 2 xAI

1,262

52%

1,262

#112

Qwen 2.5 72B Alibaba

1,261

52%

1,261

#112

Command R+Cohere

1,261

52%

1,261

#112

Command A Cohere

1,261

52%

1,261

#112

Claude 3 Haiku Anthropic

1,261

52%

1,261

#116

Phi-4 Microsoft

1,256

51%

1,256

#117

GPT-4 Turbo OpenAI

1,255

51%

1,255

#118

Command R+ (08-2024)Cohere

1,250

50%

1,250

#119

Llama 3.3 70B Meta

1,243

49%

1,243

#120

Claude Haiku 4.5 Anthropic

1,240

49%

1,240

#121

Claude 3 Opus Anthropic

1,232

47%

1,232

#122

Llama 3.1 405B Meta

1,229

47%

1,229

#123

GPT-4o mini OpenAI

1,222

46%

1,222

#124

Llama 3.1 8B Instruct Meta

1,211

44%

1,211

#125

Llama 3.1 70B Meta

1,198

43%

1,198

#126

Claude 3.5 Haiku Anthropic

1,178

40%

1,178

#127

Llama 3.2 3B Instruct Meta

1,166

38%

1,166

#128

Mixtral 8x22B Mistral AI

1,146

35%

1,146

#129

Llama 3.2 1B Instruct Meta

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 Fable 5 currently holds the top score on the Arena Elo benchmark. See our full rankings table above for the complete leaderboard with 129 models.

We update benchmark data from multiple sources including HuggingFace open-source model leaderboards 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