local/sefi-image-5b-rl Benchmark — 48/100 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.

MakerSefi

FamilySefi Image

Model size5B5B params

Costself-hostedfree

Latency131s median

Run targetlocal/sefi-image-5b-rl

Effective requestmodel: sefi-image-5b-rl · 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

58%

Typography Style67%

Writing accuracy50%

Spatial Reasoning

49%

Attributes Binding33%

Compositionality56%

Counting44%

Negation33%

Relative Position42%

Scale & Proportions44%

Human realism

49%

Faces & Expressions42%

Full Body33%

Hands42%

Multi-Subject50%

Professional Studio

48%

Camera & Lighting42%

Color Precision42%

Photorealism33%

Graphical design

69%

Data Visualisation67%

Layout & Design56%

Style Diversity58%

Truthfulness

66%

Photorealism33%

Physics & Reflections67%

World Knowledge75%