Source: https://huggingface.co/6chan/flux1-kontext-dev-fp8

A combined Flux Kontext model, quantized in NF4, for weak video cards. No need for clip, t5, vae, everything is combined.

Based on the fp8 version of the Flux Kontext model.

Combined with standard t5, clip, vae, for regular Flux.

The model is designed for those with a weak video card, like me (1070), or for those with little RAM. NF4 often showed better results than GGUF.