우리들의 비전

LiveBench 에이전틱 코딩

이 벤치마크는?

LiveBench 에이전틱 코딩 카테고리 점수 (0~100). 자율 코드 생성·수정·테스트 등 멀티스텝 코딩 능력을 평가합니다.

출처: LiveBench
순위모델
#1

OpenAI

GPT-5.4

70.0
#2

Google

Gemini 3.1 Pro

65.0
#3

Anthropic

Claude Sonnet 4.6

63.3
#4

Anthropic

Claude Opus 4.6

61.7
#5

Anthropic

Claude Opus 4.7

60.0
#6

Moonshot AI

Kimi K2.6

58.3
#7

OpenAI

GPT-5.5

56.7
#8

Z.ai

GLM-5

55.0
#9

Z.ai

GLM-5.1

55.0
#10

Alibaba

Qwen3.6 Plus

55.0
#11

OpenAI

GPT-5

51.7
#12

MiniMax

MiniMax M2.5

51.7
#13

Anthropic

Claude Opus 4.5

50.0
#14

MiniMax

MiniMax M2.7

50.0
#15

Anthropic

Claude Opus 4.1

48.3
#16

Anthropic

Claude Sonnet 4.5

48.3
#17

Moonshot AI

Kimi K2.5

48.3
#18

DeepSeek

DeepSeek V3.2

46.7
#19

Google

Gemini 3 Flash

43.3
#20

Grok

Grok 4.20 (Reasoning)

43.3
#21

Anthropic

Claude Sonnet 4

40.0
#22

Google

Gemma 4 31B

40.0
#23

Grok

Grok 4.20

38.3
#24

OpenAI

GPT-5 Mini

35.0
#25

Anthropic

Claude Haiku 4.5

33.3
#26

Google

Gemini 2.5 Pro

33.3
#27

Google

Gemini 3.1 Flash Lite

33.3
#28

Grok

Grok 4.1 Fast (Reasoning)

31.7
#29

Xiaomi

MiMo-V2-Pro

30.0
#30

OpenAI

GPT-5 Nano

28.3
#31

OpenAI

GPT-5.4 Nano

27.1
#32

NVIDIA

Nemotron 3 Super

23.0
#33

OpenAI

GPT-5.4 Mini

17.0
#34

Google

Gemini 2.5 Flash

16.7
#35

OpenAI

GPT OSS 120B

16.7
#36

Grok

Grok 4.1 Fast

10.0
#37

Google

Gemini 2.5 Flash Lite

5.0
#38

Arcee AI

Trinity Large Thinking

3.3