Qwen 2.5 VL 7B vs.Qwen 3.5 122B (A10B)
Passed:
Qwen 2.5 VL 7B 55.6% (45/81)
Qwen 3.5 122B (A10B) 78.0% (64/82)
Average request time:
Qwen 2.5 VL 7B 45.63s
Qwen 3.5 122B (A10B) 1.54s
Passed:
Qwen 2.5 VL 7B 55.6% (45/81)
Qwen 3.5 122B (A10B) 78.0% (64/82)
Average request time:
Qwen 2.5 VL 7B 45.63s
Qwen 3.5 122B (A10B) 1.54s
Summary
Annotation Understanding
Qwen 2.5 VL 7B
1 / 2
Qwen 3.5 122B (A10B)
1 / 2
CAPTCHA
Qwen 2.5 VL 7B
2 / 2
Qwen 3.5 122B (A10B)
2 / 2
Color Identification
Qwen 2.5 VL 7B
0 / 1
Qwen 3.5 122B (A10B)
1 / 1
Defect Detection
Qwen 2.5 VL 7B
9 / 15
Qwen 3.5 122B (A10B)
13 / 15
Document Understanding
Qwen 2.5 VL 7B
8 / 10
Qwen 3.5 122B (A10B)
9 / 10
Localization
Qwen 2.5 VL 7B
1 / 1
Qwen 3.5 122B (A10B)
1 / 1
OCR
Qwen 2.5 VL 7B
6 / 9
Qwen 3.5 122B (A10B)
7 / 9
Object Counting
Qwen 2.5 VL 7B
1 / 9
Qwen 3.5 122B (A10B)
5 / 11
Object Detection
Qwen 2.5 VL 7B
1 / 2
Qwen 3.5 122B (A10B)
2 / 2
Object Measurement
Qwen 2.5 VL 7B
0 / 1
Qwen 3.5 122B (A10B)
1 / 1
Object Understanding
Qwen 2.5 VL 7B
5 / 11
Qwen 3.5 122B (A10B)
8 / 11
Receipt Reading
Qwen 2.5 VL 7B
0 / 1
Qwen 3.5 122B (A10B)
1 / 1
Sign Understanding
Qwen 2.5 VL 7B
2 / 2
Qwen 3.5 122B (A10B)
2 / 2
Spatial Relations
Qwen 2.5 VL 7B
11 / 17
Qwen 3.5 122B (A10B)
14 / 17
Web Action Understanding
Qwen 2.5 VL 7B
4 / 4
Qwen 3.5 122B (A10B)
4 / 4
Contribute a Prompt
Have an idea for a prompt? Open a pull request on the project repository!