Hi-Dev

Initial release of Krea 2 Turbo quantized in NVFP4 format.

This is an ultra-low 4-bit precision scaling tailored for compatible NVIDIA architectures (optimized for RTX 50xx Blackwell series) to drastically reduce VRAM utilization while preserving generation speeds and semantic accuracy.

Recommended Inference Settings:

* Steps: 8 (Distilled Turbo model)

* CFG: 0.0 (Disabled)

* Timestep-shift Mu: 1.15

* Resolution: 1024px – 2048px

* Text Encoder: Requires Qwen3-VL 4B text encoder (set type to 'krea2' in ComfyUI CLIPLoader)

Derived from Krea 2 developed by Krea.ai. Subject to the Krea 2 Community License Agreement.

[NOTE: This is a standalone Krea 2 Turbo DiT model, uploaded under the FLUX.1 category due to architectural similarity and platform limitations.]