🎨 Quantized weights in GGUF format have been released for the Krea 2 (v1.0) model — a high-performance Diffusion Transformer (DiT) architecture with 12–13 billion parameters.
Both Krea 2 Base and the optimized Krea 2 Turbo variants are available, allowing image generation to be run on consumer hardware with low VRAM via ComfyUI.
🌍 The transition of large image generation models (DiT) to the GGUF format expands the possibilities for local use of Krea.ai-level models, lowering the barrier to entry for developers and researchers who do not have access to powerful server GPUs.
👤 You can now run the powerful Krea 2 neural network directly on your home PC via ComfyUI. The quantized versions (GGUF) occupy significantly less video memory while maintaining high generation quality.
Source 1: https://huggingface.co/realrebelai/KREA-2_GGUFs Source 2: https://huggingface.co/vantagewithai/Krea-2-Turbo-GGUF