Krea 2 is Krea AI's first image foundation model, built entirely from scratch, with a focus on aesthetic diversity and creative control. According to its technical report (released on June 23, 2026), it is a 12.9-billion-parameter single-stream diffusion Transformer that uses Qwen3-VL as its text encoder, combining text-to-image generation with prompt expansion and visual style references. This website, krea-2.pro, is an independent workspace, not affiliated with, endorsed by, or sponsored by Krea AI. It references Krea and Krea 2 only to identify the model and related workflows. The model facts on the pages come from Krea's official technical report, product documentation, and public model cards.
Krea 2
Krea 2 is the first image foundation model built from scratch by Krea AI, released in 2026. The official technical report describes it as a 12.9-billion-parameter single-stream diffusion Transformer that uses Qwen3-VL as its text encoder, emphasizes aesthetic diversity and creative control, and combines text-to-image generation with prompt expansion and visual style references. This website, krea-2.pro, is an independent workspace, not an official Krea AI site. Around Krea 2 it offers processes such as exploratory text-to-image generation, style references, mood boards, creativity adjustment, and rapid visual direction iteration, and features annual credit packages.
