Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

MMLU Pro

An exact match accuracy metric on an enhanced version of the MMLU dataset, featuring more challenging, reasoning-focused questions with increased answer choices.
Source:

Model Performance

#1
Gemini 3.1 Pro Preview
89.0%
#2
Claude Opus 4.6
88.3%
#3
Claude Opus 4.6 (Thinking)
87.9%
#4
Gemini 3.0 Flash
87.3%
#5
GPT-5.4
87.1%
#6
Claude 4.1 Opus (Thinking)
86.6%
#7
Claude Sonnet 4.5 (Thinking)
86.0%
#8
Claude Sonnet 4.6
85.9%
#9
Claude 4.0 Opus (Thinking)
85.9%
#10
Kimi K2.5
85.7%
#11
GLM-5
85.6%
#12
GPT-5
85.0%
#13
GPT-5 (Thinking)
85.0%
#14
Gemini 2.5 Pro (Jun 2025)
84.7%
#15
Claude 4.0 Opus
84.6%
#16
Claude Sonnet 4.5
84.4%
#17
Claude 4.1 Opus
84.2%
#18
GLM-4.7
84.0%
#19
OpenAI o3 (High Effort)
84.0%
#20
OpenAI o3 (Medium Effort)
84.0%
#21
Grok 4 (Thinking)
83.7%
#22
Grok 4
83.7%
#23
DeepSeek V3.2 (Thinking)
83.2%
#24
GPT-5.4 Mini
82.8%
#25
Gemini 2.5 Pro (Thinking)
82.3%
#26
Gemini 2.5 Pro
82.3%
#27
Claude 4.0 Sonnet (Thinking)
82.0%
#28
Qwen 3 Max (Thinking)
81.7%
#29
Qwen 3 Max Preview
81.7%
#30
OpenAI o1
81.7%
#31
Gemini 2.5 Flash
81.3%
#32
DeepSeek V3.2
81.2%
#33
Claude 3.7 Sonnet (Thinking)
80.8%
#34
GPT-5 Mini
80.6%
#35
Llama 3.1 405B
80.6%
#36
GPT-5 Mini (Thinking)
80.6%
#37
Kimi K2
80.4%
#38
Grok 3 Mini
79.3%
#39
GPT-4.5
78.9%
#40
Mistral Large 3
78.9%
#41
Claude 3.7 Sonnet
78.6%
#42
OpenAI o4 Mini (Medium Effort)
78.4%
#43
OpenAI o4 Mini (High Effort)
78.4%
#44
GPT-4.1
78.3%
#45
Gemini 2.0 Pro
78.3%
#46
Qwen 3
78.2%
#47
Qwen 3 (Thinking)
78.2%
#48
Claude Haiku 4.5
77.8%
#49
Grok 3
77.7%
#50
Grok 3 (Thinking)
77.7%
#51
Gemini 2.0 Flash Thinking (Jan 2025)
77.6%
#52
Gemini 2.0 Flash (Thinking)
77.6%
#53
Claude 4.0 Sonnet
77.1%
#54
Llama 4 Maverick 17B
77.1%
#55
GPT OSS 120B
76.8%
#56
DeepSeek V3 (Mar 2025)
76.6%
#57
Claude Haiku 4.5 (Thinking)
76.4%
#58
OpenAI o3 Mini (Medium Effort)
76.3%
#59
OpenAI o3 Mini (High Effort)
76.3%
#60
Gemini 3.1 Flash Lite Preview
76.3%
#61
Claude 3.5 Sonnet
76.0%
#62
GPT-5 Nano
75.4%
#63
GPT-5 Nano (Thinking)
75.4%
#64
Gemini 2.0 Flash
74.9%
#65
GPT-5.4 Nano
74.6%
#66
Gemini 1.5 Pro
72.6%
#67
GPT-4o
72.0%
#68
Mistral Medium 3.1
71.6%
#69
OpenAI o1 Mini
71.3%
#70
DeepSeek V3
70.9%
#71
Phi-4
68.2%
#72
DeepSeek R1
67.9%
#73
Llama 3.3 70B
66.6%
#74
Mistral Large 2
66.3%
#75
Claude 3 Opus
66.2%
#76
Claude Opus 4.5
64.4%
#77
Cohere Command A
61.3%
#78
Claude 3.5 Haiku
60.1%
#79
GPT-4o Mini
58.6%
#80
Grok 4.20
55.6%
#81
Claude 3 Sonnet
53.2%
#82
Claude Opus 4.5 (Thinking)
45.2%
#83
GPT-5.2
43.8%
#84
Grok 4.20 (Reasoning)
43.6%
#85
GPT-3.5 Turbo
40.2%
#86
Cohere Command R+
37.7%
#87
Llama 2 70B
34.1%
#88
Llama 2 13B
34.0%