Arena 어려운 프롬프트 Elo

이 벤치마크는?

난이도 높은 프롬프트에서의 사용자 선호도를 집계한 Arena Elo 점수입니다.

출처: Arena Intelligence
순위모델
#1

Anthropic

Claude Opus 4.6

1535
#2

Google

Gemini 3.1 Pro

1516
#3

OpenAI

GPT-5.4

1507
#4

OpenAI

GPT-5.4 Pro

1507
#5

Grok

Grok 4.20

1505
#6

Grok

Grok 4.20 (Reasoning)

1505
#7

Anthropic

Claude Opus 4.5

1500
#8

Anthropic

Claude Sonnet 4.6

1499
#9

Z.ai

GLM-5.1

1494
#10

Google

Gemini 3 Flash

1493
#11

Anthropic

Claude Sonnet 4.5

1485
#12

OpenAI

GPT-5.4 Mini

1484
#13

Anthropic

Claude Opus 4.1

1480
#14

Xiaomi

MiMo-V2-Pro

1478
#15

Z.ai

GLM-5

1476
#16

Moonshot AI

Kimi K2.5

1475
#17

Google

Gemma 4 31B

1474
#18

Baidu

ERNIE 5.0 Thinking

1470
#19

Qwen

Qwen3.5 397B A17B

1465
#20

Meituan

Longcat Flash Chat

1464
#21

Google

Gemini 2.5 Pro

1461
#22

Anthropic

Claude Opus 4

1456
#23

Google

Gemini 3.1 Flash Lite

1448
#24

DeepSeek

DeepSeek V3.2

1447
#25

OpenAI

GPT-5

1446
#26

Grok

Grok 4.1 Fast

1442
#27

Grok

Grok 4.1 Fast (Reasoning)

1442
#28

Anthropic

Claude Haiku 4.5

1437
#29

Anthropic

Claude Sonnet 4

1431
#30

MiniMax

MiniMax M2.5

1429
#31

MiniMax

MiniMax M2.7

1426
#32

Google

Gemini 2.5 Flash

1420
#33

OpenAI

GPT-5.4 Nano

1417
#34

OpenAI

GPT-5 Mini

1402
#35

Arcee AI

Trinity Large Thinking

1400
#36

Google

Gemini 2.5 Flash Lite

1391
#37

OpenAI

GPT OSS 120B

1363
#38

Amazon

Nova 2 Lite

1360
#39

OpenAI

GPT-5 Nano

1353
#40

Meta AI

Llama 4 Maverick

1338
#41

Meta AI

Llama 4 Scout

1329
#42

OpenAI

GPT-4.1

1311