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/hidream-i1-full-17b

Local

HiDream-I1 Full is a 17B-parameter open image generation foundation model from HiDream.ai, designed for high-quality text-to-image generation with strong prompt following across styles.

Maker: HiDream.ai
Family: HiDream-I1
Model Size: 17B
high
Cost: local run; no API price
not_applicable
Run Target: gx10/hidream-i1
Effective Request: response_format: b64_json · size: 1024x1024 · num_inference_steps: 28
Sources: HiDream-I1 repository HiDream-I1-Full model card

42.9

Overall

36%

Capability

49.9

Est. Preference

69

Pass

123

Fail

91.3s

Avg Latency

90.4s

Min Latency

94.2s

Max Latency

All 192 generations

Text Rendering47%

Typography Style100%

Writing accuracy33%

Spatial Reasoning33%

Attributes Binding44%

Compositionality56%

Counting22%

Negation22%

Relative Position25%

Scale & Proportions33%

Human realism17%

Faces & Expressions42%

Full Body8%

Hands0%

Multi-Subject17%

Truthfulness33%

Photorealism0%

Physics & Reflections42%

World Knowledge33%

Professional Studio63%

Camera & Lighting67%

Color Precision58%

Photorealism67%

Graphical design42%

Data Visualisation0%

Layout & Design11%

Style Diversity75%