ImageBench

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 ImageBench V1 run.

local/sefi-image-5b-base

Local

SeFi Image 5B Base is a locally fine-tuned text-to-image model produced by the SeFi image-generation fine-tuning pipeline and run on an NVIDIA DGX Spark. It is the ~5B base (non-distilled) variant of the SeFi Image family. It is not a publicly released hosted product; no external model card or citation is disclosed.

Maker
SeFi pipeline
Family
SeFi Image
Model Size
~5B
estimated
Cost
local run; no API price
not_applicable
Run Target
gx10/sefi-image-5b-base
Effective Request
Effective request fields unknown
49.7
Overall
59%
Capability
40.4
Est. Preference
113
Pass
79
Fail
130.6s
Avg Latency
124.4s
Min Latency
376.6s
Max Latency
Text Rendering87%Spatial Reasoning68%Human realism24%Truthfulness37%Professional Studio96%Graphical design63%Preference40%Latency0%

All 192 generations

Text Rendering87%

Typography Style100%

Writing accuracy83%

Spatial Reasoning68%

Attributes Binding89%

Compositionality78%

Counting67%

Negation56%

Relative Position67%

Scale & Proportions56%

Human realism24%

Faces & Expressions50%

Full Body0%

Hands8%

Multi-Subject50%

Truthfulness37%

Photorealism33%

Physics & Reflections58%

World Knowledge17%

Professional Studio96%

Camera & Lighting92%

Color Precision100%

Photorealism100%

Graphical design63%

Data Visualisation0%

Layout & Design56%

Style Diversity83%