Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

LiveBench (Math)

Evaluates performance on recent high school math competition problems and enhanced AMPS questions from the past 12 months.
Source:

Model Performance

#1
GPT-5
92.8%
#2
Gemini 3.1 Pro Preview
91.0%
#3
Claude Opus 4.5 (Thinking)
90.4%
#4
Claude Opus 4.6 (Thinking)
89.3%
#5
Claude 4.0 Opus (Thinking)
88.2%
#6
Claude Sonnet 4.6
87.0%
#7
GPT-5 (Thinking)
86.2%
#8
GPT-5 Mini
86.0%
#9
DeepSeek R1
85.3%
#10
DeepSeek V3.2 (Thinking)
85.0%
#11
OpenAI o3 (High Effort)
85.0%
#12
OpenAI o4 Mini (High Effort)
84.9%
#13
Gemini 2.5 Flash
84.1%
#14
Gemini 2.5 Pro
83.3%
#15
GPT-5.1 Codex Max
83.2%
#16
GPT-5 Mini (Thinking)
82.2%
#17
OpenAI o4 Mini (Medium Effort)
81.0%
#18
OpenAI o3 (Medium Effort)
80.7%
#19
Claude Sonnet 4.5 (Thinking)
79.3%
#20
Claude 3.7 Sonnet (Thinking)
79.0%
#21
DeepSeek V3
78.9%
#22
Claude 4.0 Opus
78.8%
#23
Claude Haiku 4.5 (Thinking)
77.5%
#24
Grok 3 Mini
77.0%
#25
Gemini 2.5 Flash (Thinking)
75.3%
#26
Qwen 3 (Thinking)
73.4%
#27
Claude 4.1 Opus (Thinking)
73.2%
#28
Claude 4.0 Sonnet (Thinking)
70.5%
#29
GPT OSS 120B
68.9%
#30
GPT-5 Nano (Thinking)
68.4%
#31
Gemini 2.5 Pro (Thinking)
68.3%
#32
Qwen 3 Max Preview
68.0%
#33
GPT-4.5
67.9%
#34
Qwen 3
67.4%
#35
Claude Opus 4.5
66.3%
#36
Claude 3.7 Sonnet
64.7%
#37
DeepSeek V3.2
64.0%
#38
Claude 4.1 Opus
62.8%
#39
Phi-4
62.8%
#40
Claude Sonnet 4.5
62.6%
#41
GPT-4.1
62.4%
#42
Llama 4 Maverick 17B
60.6%
#43
Claude 4.0 Sonnet
60.4%
#44
Mistral Medium 3.1
59.7%
#45
GPT-5 Nano
59.0%
#46
GPT-5.2
58.2%
#47
Kimi K2
58.1%
#48
Claude Haiku 4.5
58.0%
#49
GPT-4o
55.7%
#50
Claude 3.5 Sonnet
50.5%
#51
Cohere Command A
45.5%
#52
Mistral Large
42.2%
#53
Claude 3.5 Haiku
34.8%
#54
Cohere Command R+
22.8%