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/flux-2-klein-4b

Local

The local FLUX.2 [klein] 4B run uses Black Forest Labs' compact FLUX.2 open-weight model, selected for maximum speed and deployment on modest local hardware.

Reproducibility

Download on motif64.sh ↗

You can download the exact source code and configuration that was used to generate these images, packaged as a capsule: FLUX.2 [klein] 4B (diffusers) (NVIDIA GB10, 24 GB). Run it on your own machine with:

m64 m64_17808a488aff

Maker: Black Forest Labs
Family: FLUX.2 [klein]
Model Size: 4B
high
Cost: local run; no API price
not_applicable
Run Target: gx10/flux2-klein-4b
Effective Request: Effective request fields unknown
Sources: Black Forest Labs FLUX.2 overview Black Forest Labs FLUX.2 inference repository

53.6

Overall

53%

Capability

54.6

Est. Preference

101

Pass

91

Fail

4.5s

Avg Latency

4.2s

Min Latency

5.2s

Max Latency

All 192 generations

Text Rendering40%

Typography Style100%

Writing accuracy25%

Spatial Reasoning49%

Attributes Binding78%

Compositionality67%

Counting33%

Negation11%

Relative Position50%

Scale & Proportions56%

Human realism55%

Faces & Expressions67%

Full Body25%

Hands75%

Multi-Subject50%

Truthfulness30%

Photorealism0%

Physics & Reflections50%

World Knowledge17%

Professional Studio89%

Camera & Lighting92%

Color Precision92%

Photorealism67%

Graphical design50%

Data Visualisation0%

Layout & Design22%

Style Diversity83%