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/recraft/v3

API

Recraft V3 is Recraft AI's text-to-image model, noted for accurate typography and brand-consistent raster + vector design; served here through fal.

Maker
Recraft
Family
Recraft V3
Model Size
not disclosed
unknown
Cost
$0.04 / image
high
Run Target
fal/fal-ai/recraft/v3/text-to-image
Effective Request
seed: 7
Sources
Recraft V3 on fal
45.5
Overall
37%
Capability
54.6
Est. Preference
70
Pass
122
Fail
10.1s
Avg Latency
7.6s
Min Latency
129.7s
Max Latency
Text Rendering67%Spatial Reasoning37%Human realism12%Truthfulness33%Professional Studio56%Graphical design42%Preference55%Latency32%

All 192 generations

Text Rendering67%

Typography Style100%

Writing accuracy58%

Spatial Reasoning37%

Attributes Binding56%

Compositionality67%

Counting11%

Negation22%

Relative Position33%

Scale & Proportions33%

Human realism12%

Faces & Expressions33%

Full Body0%

Hands8%

Multi-Subject0%

Truthfulness33%

Photorealism0%

Physics & Reflections25%

World Knowledge50%

Professional Studio56%

Camera & Lighting67%

Color Precision58%

Photorealism0%

Graphical design42%

Data Visualisation0%

Layout & Design33%

Style Diversity58%