Ghost Palette

local/krea-2-turbo — ImageBench V1

ImageBench V1 —
192 evaluations across 6 categories
Benchmark V1 verdicts are produced by VLM judges and can contain mistakes. Treat PASS/FAIL labels as machine-assisted assessments, and inspect the images yourself. Learn more about the methodology.

Generation Details

Source-backed model context, size, cost, and request settings for this run.

MakerKrea
FamilyKrea
Model size12B12B params
Costself-hostedfree
Latency73.2s median
Run targetlocal/krea-2-turbo
Effective requestmodel: krea-2-turbo · size: 1024x1024 · seed: 42
SourcesModel cardWeights & checkpoint
Text60%Spatial60%Human67%Pro Studio63%Graphical45%Truth52%Preference55%Latency0%

All 192 generations

Every challenge, grouped by category and subcategory. Solid-bordered tiles with a check passed the VLM judge; dashed tiles with a cross failed. Images are placeholders until generated.

Text Rendering

60%
Typography Style33%
Writing accuracy67%

Spatial Reasoning

60%
Attributes Binding44%
Compositionality67%
Counting56%
Negation44%
Relative Position67%
Scale & Proportions56%

Human realism

67%
Faces & Expressions58%
Full Body67%
Hands58%
Multi-Subject67%

Professional Studio

63%
Camera & Lighting50%
Color Precision50%
Photorealism100%

Graphical design

45%
Data Visualisation67%
Layout & Design44%
Style Diversity33%

Truthfulness

52%
Photorealism0%
Physics & Reflections58%
World Knowledge50%