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/prxpixel-t2i-7b

Local

PRXPixel is Photoroom's 7B pixel-space text-to-image model, denoising raw RGB directly without a VAE and using a Qwen3-VL text encoder for prompt conditioning.

Maker
Photoroom
Family
PRXPixel
Model Size
7B
high
Cost
local run; no API price
not_applicable
Run Target
gx10/prxpixel-t2i-7b
Effective Request
response_format: b64_json · seed: 7
Sources
PRXPixel model cardPRXPixel Diffusers docs
46.3
Overall
43%
Capability
50.0
Est. Preference
82
Pass
110
Fail
64.7s
Avg Latency
64.5s
Min Latency
65.1s
Max Latency
Text Rendering47%Spatial Reasoning49%Human realism17%Truthfulness26%Professional Studio78%Graphical design50%Preference50%Latency0%

All 192 generations

Text Rendering47%

Typography Style100%

Writing accuracy33%

Spatial Reasoning49%

Attributes Binding67%

Compositionality78%

Counting22%

Negation33%

Relative Position58%

Scale & Proportions33%

Human realism17%

Faces & Expressions33%

Full Body8%

Hands0%

Multi-Subject33%

Truthfulness26%

Photorealism33%

Physics & Reflections33%

World Knowledge17%

Professional Studio78%

Camera & Lighting75%

Color Precision83%

Photorealism67%

Graphical design50%

Data Visualisation0%

Layout & Design11%

Style Diversity92%