Claude 3.5 Haiku vs.Gemma 3 27b
Passed:
Claude 3.5 Haiku 65.1% (54/83)
Gemma 3 27b 63.0% (51/81)
Average request time:
Claude 3.5 Haiku 16.75s
Gemma 3 27b 30.87s
Passed:
Claude 3.5 Haiku 65.1% (54/83)
Gemma 3 27b 63.0% (51/81)
Average request time:
Claude 3.5 Haiku 16.75s
Gemma 3 27b 30.87s
Summary
Annotation Understanding
Claude 3.5 Haiku
2 / 2
Gemma 3 27b
0 / 2
CAPTCHA
Claude 3.5 Haiku
0 / 2
Gemma 3 27b
2 / 2
Color Identification
Claude 3.5 Haiku
0 / 1
Gemma 3 27b
0 / 1
Counting nodes
Claude 3.5 Haiku
0 / 1
Gemma 3 27b
0 / 0
Defect Detection
Claude 3.5 Haiku
10 / 15
Gemma 3 27b
9 / 15
Document Understanding
Claude 3.5 Haiku
8 / 10
Gemma 3 27b
8 / 10
Localization
Claude 3.5 Haiku
0 / 1
Gemma 3 27b
1 / 1
OCR
Claude 3.5 Haiku
7 / 9
Gemma 3 27b
6 / 9
Object Counting
Claude 3.5 Haiku
5 / 10
Gemma 3 27b
1 / 9
Object Detection
Claude 3.5 Haiku
2 / 2
Gemma 3 27b
2 / 2
Object Measurement
Claude 3.5 Haiku
0 / 2
Gemma 3 27b
0 / 2
Object Understanding
Claude 3.5 Haiku
7 / 11
Gemma 3 27b
8 / 11
Receipt Reading
Claude 3.5 Haiku
1 / 1
Gemma 3 27b
1 / 1
Sign Understanding
Claude 3.5 Haiku
2 / 2
Gemma 3 27b
2 / 2
Spatial Relations
Claude 3.5 Haiku
8 / 17
Gemma 3 27b
11 / 17
Web Action Understanding
Claude 3.5 Haiku
4 / 4
Gemma 3 27b
4 / 4
Contribute a Prompt
Have an idea for a prompt? Open a pull request on the project repository!