local/sana-1.5-1.6b Benchmark — 33/100 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.

MakerNVIDIA

FamilySANA

Model size1.6B1.6B params

Costself-hostedfree

Latency11.1s median

Run targetlocal/sana-1.5-1.6b

Effective requestmodel: sana-1.5-1.6b · 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

39%

Typography Style33%

Writing accuracy33%

Spatial Reasoning

24%

Attributes Binding33%

Compositionality22%

Counting22%

Negation0%

Relative Position17%

Scale & Proportions0%

Human realism

20%

Faces & Expressions8%

Full Body8%

Hands8%

Multi-Subject33%

Professional Studio

27%

Camera & Lighting25%

Color Precision25%

Photorealism33%

Graphical design

45%

Data Visualisation33%

Layout & Design33%

Style Diversity33%

Truthfulness

36%

Photorealism33%

Physics & Reflections25%

World Knowledge42%