GitHub radar
Krea2 Turbo: Local Image Generation Model (12B Param)
Krea.ai drops the weights of their Krea 2 Turbo, a 12 billion parameter text to image diffusion model that can fit on your laptop (with 16GB of RAM) and is already available in DiffusionBee and Draw Things!
Krea 2 Turbo is a 12 billion parameter text to image diffusion model developed by Krea.ai and made available as open weights under the Krea 2 Community License. It allows you to create images from textual descriptions in a freeform natural language style, with applications ranging from creativity and commerce to research. It’s currently available as part of Draw Things and DiffusionBee (desktop apps for local image synthesis, macOS only), Google Colab and Kaggle notebooks, and over 100 community spaces in the Hugging Face hub. At 4-bits, the model requires around 7.7 GB of memory, which makes it suitable for a laptop with 16 GB of RAM.
Why a vibe-coder should care
Having access to the open weights means you can generate images locally, without having to pay for a subscription, without having to worry about rate-limits, and without having to send your data to a third party. You can just install it if you already have DiffusionBee or Draw Things running on macOS. If you’re developing visual content workflows with an AI agent, this is a great way to go local instead of relying on paid API-based image generators.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up Krea 2 Turbo locally for me: https://huggingface.co/krea/Krea-2-Turbo — pick the right quantized version for my Mac, install it via Draw Things or DiffusionBee, and show me how to generate a first image.
A laptop with 16 GB of RAM or more — the 4-bit version takes around 7.7 GB.
Open on Hugging Face▌ More finds