GitHub radar
Krea 2: Open Source Image Model Released!
Krea AI has published the complete inference code and weights for Krea 2, a state-of-the-art text-to-image generative model which is currently the top model on the Artificial Analysis text-to-image benchmark. It comes in a Raw version suitable for fine-tuning and LoRA training, as well as a faster Turbo version that completes images in only 8 steps!
Official Inference Code & Open Weights of Krea 2, a scratch-trained text-to-image model created by Krea AI. We release both our Raw checkpoint (full-size base model for finetuning/LoRA) and our Turbo checkpoint (a distilled model that generates images within just 8 steps)! As reported by Artificial Analysis, Krea 2 is currently the state-of-the-art text-to-image model (June 2026).
Why a vibe-coder should care
Now that we have released our model weights, you can run Krea 2 locally, finetune it to your liking, or even seamlessly plug it into ComfyUI without having to deal with API queues/per-image fees. You are welcome to utilize this model for commercial purposes under the Apache 2.0 License.
▌ More finds