Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

LiveBench (Coding)

Assesses performance on LiveBench coding tasks from Leetcode and AtCoder, including both code generation and code completion challenges.
Source:

Model Performance

#1
Claude 4.0 Sonnet
80.7%
#2
GPT-5.1 Codex Max
80.7%
#3
Claude Sonnet 4.5 (Thinking)
80.4%
#4
OpenAI o4 Mini (High Effort)
80.0%
#5
Claude Opus 4.5 (Thinking)
79.7%
#6
Claude Sonnet 4.6
79.3%
#7
Claude Opus 4.5
78.5%
#8
Claude Opus 4.6 (Thinking)
78.2%
#9
OpenAI o3 (Medium Effort)
77.9%
#10
Claude 4.0 Sonnet (Thinking)
77.5%
#11
GPT-4o
77.5%
#12
OpenAI o3 (High Effort)
76.7%
#13
Gemini 3.1 Pro Preview
76.5%
#14
GPT-5.2
76.5%
#15
GPT-4.5
76.1%
#16
Claude 4.1 Opus
76.1%
#17
Claude Sonnet 4.5
76.1%
#18
DeepSeek V3.2
75.7%
#19
Gemini 2.5 Pro (Thinking)
75.7%
#20
GPT-5
75.3%
#21
Claude 4.1 Opus (Thinking)
74.7%
#22
Claude 3.7 Sonnet
74.3%
#23
Kimi K2
74.3%
#24
OpenAI o4 Mini (Medium Effort)
74.2%
#25
Claude 3.5 Sonnet
73.9%
#26
Claude 4.0 Opus
73.6%
#27
Claude 4.0 Opus (Thinking)
73.2%
#28
GPT-4.1
73.2%
#29
Claude 3.7 Sonnet (Thinking)
73.2%
#30
GPT-5 Mini
72.9%
#31
Claude Haiku 4.5 (Thinking)
72.8%
#32
Claude Haiku 4.5
72.2%
#33
GPT-5 (Thinking)
72.1%
#34
DeepSeek R1
71.4%
#35
GPT-5 Nano
70.8%
#36
Gemini 2.5 Pro
70.7%
#37
Qwen 3 Max Preview
69.6%
#38
Qwen 3 (Thinking)
69.0%
#39
DeepSeek V3
68.5%
#40
GPT-5 Mini (Thinking)
68.2%
#41
Gemini 2.5 Flash (Thinking)
67.5%
#42
Qwen 3
66.0%
#43
DeepSeek V3.2 (Thinking)
64.6%
#44
Gemini 2.5 Flash
63.5%
#45
Mistral Large
62.9%
#46
GPT-5 Nano (Thinking)
62.4%
#47
Mistral Medium 3.1
61.5%
#48
Phi-4
60.6%
#49
GPT OSS 120B
60.2%
#50
Grok 3 Mini
54.5%
#51
Cohere Command A
54.3%
#52
Llama 4 Maverick 17B
54.2%
#53
Claude 3.5 Haiku
53.2%
#54
Cohere Command R+
27.1%