Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

HumanEval

Evaluates the model's ability to generate correct Python code solutions to programming problems on the first attempt, based on function signatures and docstrings.
Source:

Model Performance

#1
GPT-5-Thinking
99.0%
#2
GPT-5
99.0%
#3
Gemini-2.5-Pro-0325
99.0%
#4
OpenAI-O4-mini-medium
99.0%
#5
OpenAI-O4-mini-high
99.0%
#6
OpenAI-O3-medium
99.0%
#7
OpenAI-O3-high
99.0%
#8
Gemini-2.5-Pro-0605
99.0%
#9
Grok-4
98.0%
#10
Grok-3-Mini-Beta
98.0%
#11
Grok-4-Thinking
98.0%
#12
Qwen-3-Thinking
98.0%
#13
Claude-3.7-Sonnet-Thinking
98.0%
#14
DeepSeek-R1
97.0%
#15
Claude-4.0-Sonnet
97.0%
#16
OpenAI-O3-mini-medium
97.0%
#17
OpenAI-O1-1217
97.0%
#18
Claude-4.0-Opus
97.0%
#19
OpenAI-O1-mini
97.0%
#20
GPT-4.1
96.0%
#21
Gemini-2.5-Flash-Thinking
96.0%
#22
Qwen-3
96.0%
#23
Claude-3.7-Sonnet
95.0%
#24
Gemini-2.0-Pro-0121
95.0%
#25
Gemini-2.5-Flash
95.0%
#26
GPT-4o-0513
94.0%
#27
Gemini-2.0-Flash-Thinking
94.0%
#28
Kimi-K2-Instruct
93.0%
#29
Claude-3.5-Sonnet-1022
93.0%
#30
DeepSeek-V3-0324
92.0%
#31
Grok-3-Beta
91.0%
#32
DeepSeek-V3
91.0%
#33
Gemini-1.5-Pro
90.0%
#34
Mistral-Medium-3.1
90.0%
#35
Mistral-Large-2
90.0%
#36
Gemini-2.0-Flash
90.0%
#37
Llama-4-Maverick-17B
88.0%
#38
GPT-4-mini
88.0%
#39
Phi-4
87.0%
#40
Llama-3.3-70B
86.0%
#41
Claude-3.5-Haiku
86.0%
#42
Claude-3-Opus
85.0%
#43
Llama-3.1-405B
85.0%
#44
Cohere-Command-A
82.0%
#45
Mistral-Large
71.0%
#46
Claude-3-Sonnet
71.0%
#47
GPT-3.5-Turbo
70.0%
#48
Cohere-Command-R-Plus
64.0%
#49
Llama-2-70B
34.0%
#50
Llama-2-7B
13.0%
#51
Llama-2-13B
0.0%