local/nucleus-image-17b-a2b Benchmark — 37/100 Overall, 192 Prompts, Every Image Shown

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.

MakerNucleus

FamilyNucleus Image

Model size17B17B params

Costself-hostedfree

Latency39.1s median

Run targetlocal/nucleus-image-17b-a2b

Effective requestmodel: nucleus-image-17b-a2b · size: 1024x1024 · seed: 42

SourcesModel cardWeights & checkpoint

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

29%

Typography Style0%

Writing accuracy17%

Spatial Reasoning

29%

Attributes Binding22%

Compositionality33%

Counting44%

Negation22%

Relative Position25%

Scale & Proportions22%

Human realism

55%

Faces & Expressions42%

Full Body42%

Hands50%

Multi-Subject83%

Professional Studio

49%

Camera & Lighting50%

Color Precision33%

Photorealism67%

Graphical design

33%

Data Visualisation33%

Layout & Design11%

Style Diversity25%

Truthfulness

55%

Photorealism67%

Physics & Reflections67%

World Knowledge42%