GPT-5 (high reasoning) vs.Gemma 3 4B
Passed:
GPT-5 (high reasoning) 73.5% (61/83)
Gemma 3 4B 42.2% (35/83)
Average request time:
GPT-5 (high reasoning) 26.96s
Gemma 3 4B 15.86s
Passed:
GPT-5 (high reasoning) 73.5% (61/83)
Gemma 3 4B 42.2% (35/83)
Average request time:
GPT-5 (high reasoning) 26.96s
Gemma 3 4B 15.86s
Summary
Annotation Understanding
GPT-5 (high reasoning)
1 / 2
Gemma 3 4B
0 / 2
CAPTCHA
GPT-5 (high reasoning)
0 / 2
Gemma 3 4B
2 / 2
Color Identification
GPT-5 (high reasoning)
1 / 1
Gemma 3 4B
0 / 1
Counting nodes
GPT-5 (high reasoning)
0 / 1
Gemma 3 4B
0 / 1
Defect Detection
GPT-5 (high reasoning)
12 / 15
Gemma 3 4B
9 / 15
Document Understanding
GPT-5 (high reasoning)
8 / 10
Gemma 3 4B
6 / 10
Localization
GPT-5 (high reasoning)
1 / 1
Gemma 3 4B
0 / 1
OCR
GPT-5 (high reasoning)
9 / 9
Gemma 3 4B
6 / 9
Object Counting
GPT-5 (high reasoning)
3 / 10
Gemma 3 4B
0 / 10
Object Detection
GPT-5 (high reasoning)
1 / 2
Gemma 3 4B
2 / 2
Object Measurement
GPT-5 (high reasoning)
0 / 2
Gemma 3 4B
0 / 2
Object Understanding
GPT-5 (high reasoning)
9 / 11
Gemma 3 4B
5 / 11
Receipt Reading
GPT-5 (high reasoning)
1 / 1
Gemma 3 4B
0 / 1
Sign Understanding
GPT-5 (high reasoning)
2 / 2
Gemma 3 4B
2 / 2
Spatial Relations
GPT-5 (high reasoning)
15 / 17
Gemma 3 4B
4 / 17
Web Action Understanding
GPT-5 (high reasoning)
4 / 4
Gemma 3 4B
2 / 4
Contribute a Prompt
Have an idea for a prompt? Open a pull request on the project repository!