FLUX.2 [klein] 4B is a fast image AI model dropped by Black Forest Labs in January 2026. It's built to run quick even on basic hardware. The model turns text into pictures and also edits images using written instructions.
This version has about 4 billion settings and runs in 4 steps, so it's much faster than bigger models. On high-end GPUs like the RTX 5090 it works in under half a second. Even on something like a 3090 or 4070 it stays around a second.
It uses a flow transformer design that helps it do both generation and editing. You can feed it one or more images as a guide.
It runs locally or through APIs. You can use it in tools like ComfyUI, Hugging Face, or Replicate. It's also open source under Apache 2.0, so you can use it for anything, even paid projects. Needs around 8.4 GB of VRAM, so it works fine on many consumer-level GPUs.
It fits well in real-time apps like live previews or creative tools. Artists and devs use it for fast edits, live workflows, or just building stuff locally.

If you'd like to access this model, you can explore the following possibilities:
Tutorial
A quick introduction to FLUX.2 [klein] 4B and how to run text-to-image and image editing workflows in ComfyUI.