Qwen 3.5 35B (A3B) vs.Cosmos Reason2 2B
Passed:
Qwen 3.5 35B (A3B) 65.9% (54/82)
Cosmos Reason2 2B 41.5% (34/82)
Average request time:
Qwen 3.5 35B (A3B) 1.28s
Cosmos Reason2 2B 1.94s
Passed:
Qwen 3.5 35B (A3B) 65.9% (54/82)
Cosmos Reason2 2B 41.5% (34/82)
Average request time:
Qwen 3.5 35B (A3B) 1.28s
Cosmos Reason2 2B 1.94s
Summary
Annotation Understanding
Qwen 3.5 35B (A3B)
1 / 2
Cosmos Reason2 2B
0 / 2
CAPTCHA
Qwen 3.5 35B (A3B)
2 / 2
Cosmos Reason2 2B
2 / 2
Color Identification
Qwen 3.5 35B (A3B)
0 / 1
Cosmos Reason2 2B
1 / 1
Defect Detection
Qwen 3.5 35B (A3B)
12 / 15
Cosmos Reason2 2B
9 / 15
Document Understanding
Qwen 3.5 35B (A3B)
8 / 10
Cosmos Reason2 2B
4 / 10
Localization
Qwen 3.5 35B (A3B)
0 / 1
Cosmos Reason2 2B
0 / 1
OCR
Qwen 3.5 35B (A3B)
8 / 9
Cosmos Reason2 2B
6 / 9
Object Counting
Qwen 3.5 35B (A3B)
2 / 11
Cosmos Reason2 2B
1 / 11
Object Detection
Qwen 3.5 35B (A3B)
0 / 2
Cosmos Reason2 2B
1 / 2
Object Measurement
Qwen 3.5 35B (A3B)
0 / 1
Cosmos Reason2 2B
0 / 1
Object Understanding
Qwen 3.5 35B (A3B)
7 / 11
Cosmos Reason2 2B
6 / 11
Receipt Reading
Qwen 3.5 35B (A3B)
1 / 1
Cosmos Reason2 2B
1 / 1
Sign Understanding
Qwen 3.5 35B (A3B)
2 / 2
Cosmos Reason2 2B
2 / 2
Spatial Relations
Qwen 3.5 35B (A3B)
10 / 17
Cosmos Reason2 2B
4 / 17
Web Action Understanding
Qwen 3.5 35B (A3B)
4 / 4
Cosmos Reason2 2B
0 / 4
Contribute a Prompt
Have an idea for a prompt? Open a pull request on the project repository!