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.

fal/qwen-image-2-pro

API

Qwen Image 2.0 Pro is Alibaba's Qwen (Tongyi Lab) flagship hosted text-to-image model, a top-10 model on the Artificial Analysis image arena known for best-in-class typography, infographics, and poster/layout rendering. Distinct from the older local Qwen-Image checkpoint on this board; released ~April 2026.

Maker
Alibaba (Qwen)
Family
Qwen Image 2.0
Model Size
not disclosed
unknown
Cost
$0.075 / image
high
Run Target
fal/fal-ai/qwen-image-2/pro/text-to-image
Effective Request
image_size: square_hd · seed: 7
66.9
Overall
71%
Capability
63.1
Est. Preference
136
Pass
56
Fail
12.8s
Avg Latency
11.0s
Min Latency
26.0s
Max Latency
Text Rendering100%Spatial Reasoning75%Human realism38%Truthfulness59%Professional Studio93%Graphical design88%Preference63%Latency25%

All 192 generations

Text Rendering100%

Typography Style100%

Writing accuracy100%

Spatial Reasoning75%

Attributes Binding89%

Compositionality89%

Counting67%

Negation89%

Relative Position83%

Scale & Proportions33%

Human realism38%

Faces & Expressions92%

Full Body17%

Hands0%

Multi-Subject50%

Truthfulness59%

Photorealism33%

Physics & Reflections50%

World Knowledge75%

Professional Studio93%

Camera & Lighting83%

Color Precision100%

Photorealism100%

Graphical design88%

Data Visualisation67%

Layout & Design78%

Style Diversity100%