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.

55.3
Overall
54%
Capability
57.1
Est. Preference
103
Pass
89
Fail
4.1s
Avg Latency
2.6s
Min Latency
11.6s
Max Latency
Text Rendering40%Spatial Reasoning68%Human realism29%Truthfulness26%Professional Studio93%Graphical design58%Preference57%Latency59%

All 192 generations

Text Rendering40%

Typography Style67%

Writing accuracy33%

Spatial Reasoning68%

Attributes Binding100%

Compositionality78%

Counting33%

Negation56%

Relative Position75%

Scale & Proportions67%

Human realism29%

Faces & Expressions67%

Full Body8%

Hands0%

Multi-Subject50%

Truthfulness26%

Photorealism33%

Physics & Reflections33%

World Knowledge17%

Professional Studio93%

Camera & Lighting92%

Color Precision100%

Photorealism67%

Graphical design58%

Data Visualisation0%

Layout & Design33%

Style Diversity92%