Gemma 4 31B vs.Qwen 2.5 VL 7B
Passed:
Gemma 4 31B 72.2% (39/54)
Qwen 2.5 VL 7B 55.6% (45/81)
Average request time:
Gemma 4 31B 15.30s
Qwen 2.5 VL 7B 45.63s
Passed:
Gemma 4 31B 72.2% (39/54)
Qwen 2.5 VL 7B 55.6% (45/81)
Average request time:
Gemma 4 31B 15.30s
Qwen 2.5 VL 7B 45.63s
Summary
Annotation Understanding
Gemma 4 31B
0 / 1
Qwen 2.5 VL 7B
1 / 2
CAPTCHA
Gemma 4 31B
2 / 2
Qwen 2.5 VL 7B
2 / 2
Color Identification
Gemma 4 31B
0 / 0
Qwen 2.5 VL 7B
0 / 1
Defect Detection
Gemma 4 31B
6 / 8
Qwen 2.5 VL 7B
9 / 15
Document Understanding
Gemma 4 31B
7 / 7
Qwen 2.5 VL 7B
8 / 10
Localization
Gemma 4 31B
0 / 1
Qwen 2.5 VL 7B
1 / 1
OCR
Gemma 4 31B
4 / 4
Qwen 2.5 VL 7B
6 / 9
Object Counting
Gemma 4 31B
2 / 8
Qwen 2.5 VL 7B
1 / 9
Object Detection
Gemma 4 31B
1 / 2
Qwen 2.5 VL 7B
1 / 2
Object Measurement
Gemma 4 31B
0 / 0
Qwen 2.5 VL 7B
0 / 1
Object Understanding
Gemma 4 31B
4 / 7
Qwen 2.5 VL 7B
5 / 11
Receipt Reading
Gemma 4 31B
1 / 1
Qwen 2.5 VL 7B
0 / 1
Sign Understanding
Gemma 4 31B
1 / 1
Qwen 2.5 VL 7B
2 / 2
Spatial Relations
Gemma 4 31B
10 / 13
Qwen 2.5 VL 7B
11 / 17
Web Action Understanding
Gemma 4 31B
4 / 4
Qwen 2.5 VL 7B
4 / 4
Contribute a Prompt
Have an idea for a prompt? Open a pull request on the project repository!