Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

LiveBench (Agentic Coding)

Evaluates the model's ability to operate in a real development environment to solve real-world repository issues.
Source:

Model Performance

#1
GPT-5
43.3%
#2
OpenAI-O3-high
36.7%
#3
Claude-4.1-Opus-Thinking
33.3%
#4
Claude-4.0-Opus-Thinking
33.3%
#5
Claude-4.0-Opus
31.7%
#6
Claude-4.0-Sonnet-Thinking
30.0%
#7
OpenAI-O4-mini-high
28.3%
#8
GPT-5-mini
28.3%
#9
OpenAI-O3-medium
28.3%
#10
DeepSeek-R1
26.7%
#11
Claude-3.7-Sonnet-Thinking
25.0%
#12
Claude-4.0-Sonnet
25.0%
#13
OpenAI-O4-mini-medium
21.7%
#14
Claude-3.7-Sonnet
21.7%
#15
DeepSeek-V3
21.7%
#16
Kimi-K2-Instruct
20.0%
#17
Gemini-2.5-Flash
18.3%
#18
GPT-5-nano
18.3%
#19
GPT-4.5
15.0%
#20
Claude-3.5-Sonnet-1022
15.0%
#21
Mistral-Medium-3.1
15.0%
#22
Grok-3-Mini-Beta
15.0%
#23
GPT-4.1
13.3%
#24
Gemini-2.5-Pro-0325
13.3%
#25
GPT-4o-0513
13.3%
#26
GPT-OSS-120B
10.0%
#27
Qwen-3
10.0%
#28
Claude-3.5-Haiku
6.7%
#29
Gemini-2.5-Flash-Thinking
6.7%
#30
Cohere-Command-A
5.0%
#31
Phi-4
5.0%
#32
Llama-4-Maverick-17B
3.3%
#33
Cohere-Command-R-Plus
1.7%
#34
Mistral-Large
1.7%