Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

LiveBench (Average)

Measures the model's overall performance across all LiveBench tasks, providing a comprehensive assessment of capabilities across reasoning, coding, math, data analysis, language, and instruction following.
Source:

Model Performance

#1
Gemini 3.1 Pro Preview
79.9%
#2
GPT-5
78.6%
#3
Claude Opus 4.6 (Thinking)
76.3%
#4
Claude Opus 4.5 (Thinking)
76.0%
#5
Claude Sonnet 4.6
75.5%
#6
OpenAI o3 (High Effort)
74.6%
#7
GPT-5.1 Codex Max
74.0%
#8
Claude 4.0 Opus (Thinking)
72.9%
#9
OpenAI o3 (Medium Effort)
72.0%
#10
OpenAI o4 Mini (High Effort)
71.5%
#11
GPT-5 Mini
70.7%
#12
GPT-5 (Thinking)
70.5%
#13
DeepSeek R1
70.1%
#14
Gemini 2.5 Pro
69.4%
#15
Claude Sonnet 4.5 (Thinking)
68.2%
#16
Claude 3.7 Sonnet (Thinking)
67.4%
#17
OpenAI o4 Mini (Medium Effort)
66.9%
#18
Claude 4.0 Opus
65.9%
#19
GPT-5 Mini (Thinking)
65.9%
#20
Gemini 2.5 Flash
64.4%
#21
DeepSeek V3
62.5%
#22
Grok 3 Mini
62.4%
#23
DeepSeek V3.2 (Thinking)
62.2%
#24
Claude 4.1 Opus (Thinking)
61.8%
#25
Claude Haiku 4.5 (Thinking)
61.3%
#26
Claude 4.0 Sonnet (Thinking)
61.3%
#27
Claude Opus 4.5
59.1%
#28
GPT-4.5
58.7%
#29
Claude 3.7 Sonnet
58.5%
#30
Gemini 2.5 Pro (Thinking)
58.3%
#31
GPT-5 Nano
56.5%
#32
GPT-4.1
55.9%
#33
GPT-4o
54.7%
#34
Claude 4.1 Opus
54.4%
#35
Claude Sonnet 4.5
53.7%
#36
Gemini 2.5 Flash (Thinking)
53.1%
#37
Qwen 3 (Thinking)
53.0%
#38
DeepSeek V3.2
51.8%
#39
Claude 3.5 Sonnet
51.8%
#40
Claude 4.0 Sonnet
51.0%
#41
Mistral Medium 3.1
50.6%
#42
Phi-4
49.3%
#43
GPT-5.2
48.9%
#44
Qwen 3 Max Preview
48.8%
#45
GPT-5 Nano (Thinking)
48.6%
#46
Kimi K2
48.1%
#47
Llama 4 Maverick 17B
47.8%
#48
GPT OSS 120B
46.1%
#49
Claude Haiku 4.5
45.3%
#50
Cohere Command A
44.2%
#51
Qwen 3
43.6%
#52
Mistral Large
43.3%
#53
Claude 3.5 Haiku
39.5%
#54
Cohere Command R+
30.1%