Arena 지시 따르기 Elo

이 벤치마크는?

지시 따르기 프롬프트에서의 사용자 선호도를 집계한 Arena Elo 점수입니다.

출처: Arena Intelligence
순위모델
#1

Anthropic

Claude Opus 4.6

1512
#2

Google

Gemini 3.1 Pro

1490
#3

OpenAI

GPT-5.4

1486
#4

OpenAI

GPT-5.4 Pro

1486
#5

Anthropic

Claude Opus 4.5

1485
#6

Anthropic

Claude Sonnet 4.6

1478
#7

Grok

Grok 4.20

1465
#8

Grok

Grok 4.20 (Reasoning)

1465
#9

Anthropic

Claude Sonnet 4.5

1463
#10

Z.ai

GLM-5.1

1461
#11

Anthropic

Claude Opus 4.1

1459
#12

Google

Gemini 3 Flash

1459
#13

OpenAI

GPT-5.4 Mini

1453
#14

Google

Gemma 4 31B

1452
#15

Z.ai

GLM-5

1448
#16

Xiaomi

MiMo-V2-Pro

1448
#17

Google

Gemini 2.5 Pro

1442
#18

Anthropic

Claude Opus 4

1442
#19

Moonshot AI

Kimi K2.5

1442
#20

Baidu

ERNIE 5.0 Thinking

1432
#21

Qwen

Qwen3.5 397B A17B

1431
#22

DeepSeek

DeepSeek V3.2

1420
#23

Meituan

Longcat Flash Chat

1414
#24

Anthropic

Claude Sonnet 4

1414
#25

Google

Gemini 3.1 Flash Lite

1412
#26

Anthropic

Claude Haiku 4.5

1411
#27

OpenAI

GPT-5

1409
#28

Google

Gemini 2.5 Flash

1403
#29

MiniMax

MiniMax M2.5

1403
#30

Grok

Grok 4.1 Fast

1400
#31

Grok

Grok 4.1 Fast (Reasoning)

1400
#32

MiniMax

MiniMax M2.7

1392
#33

OpenAI

GPT-5.4 Nano

1388
#34

OpenAI

GPT-5 Mini

1374
#35

Arcee AI

Trinity Large Thinking

1372
#36

Google

Gemini 2.5 Flash Lite

1365
#37

Amazon

Nova 2 Lite

1332
#38

OpenAI

GPT OSS 120B

1326
#39

OpenAI

GPT-5 Nano

1325
#40

Meta AI

Llama 4 Maverick

1313
#41

Meta AI

Llama 4 Scout

1298
#42

OpenAI

GPT-4.1

1293