Models / Regen family
Three models. One creative standard.
From fast exploration to brand-trained precision, every Regen model shares the same visual intelligence — tuned for typography, prompt fidelity, and production-ready output.
The lineup
Pick the model that fits the work.
Flagship · New
Regen 2.0
Our most capable model. Industry-leading prompt fidelity, crystal-clear typography, native transparency and reliable multi-step editing.
Fast · Efficient
Regen 1.0
The proven workhorse. Rapid generations at lower cost — ideal for ideation, moodboards and high-volume variations.
Private · Brand-trained
Custom Image Models
Fine-tune Regen on your products, characters and visual language so every output looks unmistakably yours.
Compare
Side by side.
Capability
Regen 2.0
Regen 1.0
Custom
Max resolution
4096 px
2048 px
4096 px
Text rendering
Advanced
Standard
Advanced
Transparent output
Native
—
Native
Inpaint & outpaint
Yes
Yes
Yes
Avg. generation time
~6 s
~3 s
~7 s
Open weights
Yes
Yes
Private
Brand fine-tuning
Optional
—
Included
Under the hood
Built on open foundations.
Regen models are trained on licensed and opt-in data, evaluated by working designers, and released with open weights so teams can inspect, adapt and deploy them anywhere.
01
Parameters
12B — A dense diffusion transformer designed for detail and layout control.
02
Prompt adherence
94% — Measured on our internal 2,000-prompt creative benchmark.
03
Native resolution
4K — Generate print-ready assets without separate upscaling passes.
04
Languages
38 — Accurate in-image typography across Latin, Cyrillic, Greek and more.
Start creating with Regen 2.0
Free to try. Upgrade when you are ready for higher limits, private models and API access.