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.

local/krea-2-raw

Local

Krea 2 Raw is Krea AI's undistilled base checkpoint for Krea 2, intended for fine-tuning, post-training, research, and LoRA training rather than the fastest final inference path.

Maker
Krea AI
Family
Krea 2 Open-Source
Model Size
12B
high
Cost
local run; no API price
not_applicable
Run Target
gx10/krea2-raw
Effective Request
response_format: b64_json · size: 1024x1024 · num_inference_steps: 52 · guidance_scale: 3.5 · negative_prompt: blurry, low quality, distorted, deformed, low resolution
Sources
Krea 2 Open-Source pageKrea 2 Raw model cardKrea 2 Turbo model card
42.1
Overall
57%
Capability
26.9
Est. Preference
110
Pass
82
Fail
912.2s
Avg Latency
905.6s
Min Latency
938.3s
Max Latency
Text Rendering73%Spatial Reasoning75%Human realism14%Truthfulness44%Professional Studio82%Graphical design67%Preference27%Latency0%

All 192 generations

Text Rendering73%

Typography Style100%

Writing accuracy67%

Spatial Reasoning75%

Attributes Binding100%

Compositionality89%

Counting67%

Negation67%

Relative Position92%

Scale & Proportions33%

Human realism14%

Faces & Expressions33%

Full Body0%

Hands8%

Multi-Subject17%

Truthfulness44%

Photorealism33%

Physics & Reflections50%

World Knowledge42%

Professional Studio81%

Camera & Lighting83%

Color Precision92%

Photorealism33%

Graphical design67%

Data Visualisation0%

Layout & Design56%

Style Diversity92%