LocalAI

See the speed before choosing hardware

Experience creation time for each device and play sample

Image/video creation performance

Choose your device and model and see the results for yourself.

Create image

RTX 5060 Ti 16GB · FLUX.2 Klein 4B
1024×1024 · 1 length

Shared sample · Not an output from the selected model.

Estimated: about 3 s

Things you can do locally

When pressed, the model and equipment conditions are adjusted together.

Expand settings 1024×1024 · 1 length · general
Output size
at once
Creation method

Text-to-image · Distilled model, 4 steps

Results per hour
874–1749 images
Energy at rated power
0.1–0.4 Wh
running status
Memory loadable

FLUX.2 [klein] 4B

Quick drafts, thumbnails, and iteration creation

Text → Imageimage editingApache 2.0

This is a small and fast model. Complex text and precise editing may be at a disadvantage compared to larger models.

minimal run8GB level
Recommended configuration12GB level
Image inputsupport

If you keep doing this

You can create it locally at RTX 5060 Ti 16GB

1024×1024 · 1 length standard approx. 2–4 s. Memory loadable It's a configuration.

Creation time for each device

This is a relative comparison using the same settings.

The shorter the faster
equipmentrunning statusestimated timerelative lengthwattage
selectMemory loadable2–4 s
0.1–0.4 Wh
Memory loadable1–2 s
0.2–0.3 Wh
Memory loadable0.7–1 s
0.1–0.2 Wh
Memory loadable0.4–0.8 s
0.1–0.2 Wh
Memory loadable2–3 s
0.2–0.4 Wh
Memory loadable0.5–0.9 s
0.1–0.2 Wh
Memory loadable1–2 s
0.2–0.3 Wh
Memory loadable0.4–0.8 s
0.1–0.2 Wh
Memory loadable2–3 s
0.2–0.4 Wh
Memory loadable2–4 s
0.1–0.3 Wh
Memory loadable2–4 s
0.1–0.3 Wh
No official execution pathno comparison——

Generation times are estimates scaled from reference ranges for the selected hardware and output settings. Runtime and offloading affect results; configurations without enough timing evidence show no estimate. Energy figures use rated power, not measured consumption.