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.

replicate/luma/photon

API

Photon is Luma AI's high-efficiency text-to-image model, positioned for strong prompt adherence and natural-language understanding with character consistency and multi-image reference support. It targets photorealistic generation at lower cost than prior-generation models.

Maker
Luma AI
Family
Photon
Model Size
not disclosed
unknown
Cost
$0.03 / image
high
Run Target
replicate/luma/photon
Effective Request
aspect_ratio: 1:1 · seed: 7
42.6
Overall
52%
Capability
33.7
Est. Preference
99
Pass
93
Fail
11.1s
Avg Latency
10.0s
Min Latency
16.0s
Max Latency
Text Rendering53%Spatial Reasoning49%Human realism33%Truthfulness44%Professional Studio78%Graphical design67%Preference34%Latency29%

All 192 generations

Text Rendering53%

Typography Style67%

Writing accuracy50%

Spatial Reasoning49%

Attributes Binding67%

Compositionality78%

Counting33%

Negation44%

Relative Position42%

Scale & Proportions33%

Human realism33%

Faces & Expressions42%

Full Body0%

Hands58%

Multi-Subject33%

Truthfulness44%

Photorealism33%

Physics & Reflections50%

World Knowledge42%

Professional Studio78%

Camera & Lighting75%

Color Precision75%

Photorealism100%

Graphical design67%

Data Visualisation0%

Layout & Design56%

Style Diversity92%