Project: GGUF presents Vintage Beauties Flux 2 Klein 9b
Please visit the core model page for ongoing discussions and feedback. Please support the authors there!
Brought to you by the Project: GGUF Tools
The goal of Project: GGUF is to make image generation easier on gamer/consumer hardware. We are currently focusing on Flux 2 Klein models, and have successfully converted and run many 9b models at Q5_K_M on a 3070ti Mobile (laptop) with 8gb VRAM and 32gb RAM, also Q8_0 4b models! That's both converted AND used the models on the hardware.
This is a GGUF version of the model to run on lower end hardware, that's it - nothing more. I take no credit for the model or the hard work others have put into the training or curation.
If you are the author of an original model I have uploaded and want it taken down, please get in touch! Would love to see ongoing support for the low-end hardware :)
Description
FAQ
Comments (9)
If you're looking for the GGUF version, this is it right here. Fully approved. Cheers!
Thanks for the stamp of approval :)
it would be a lot more useful if all your gguf conversions were under one primary page instead of using separate pages for each.
In theory, yes. Though it becomes a LOT harder to manage requests, updates or takedowns, etc.
workflow? does this go in Unet folder?
I use Stability Matrix, it goes in my Models/DiffusionModel - this is a base model, like plain Flux 2 Klein 9b. It should be used the same as you might with an Unsloth GGUF model.
Downloaded the Q5, really good for edit. Thank you
Yeah! When I was using Flux Klein heavily, the Q5 of vintage delivered STRONG skin/body matching edits :)
@xFennec777 it's like magic. i also use Krea edit even tho it's almost perfect but it changes alot of things and quality of output turns pixalated sometimes (after comfyui latest update). Klein is perfect for changing dress etc... with refrence image. 4b is really fast (20sec) but with errors, 9b slower (50sec) but the quality is worth it. Krea 2 slower than the two (+90sec) but when it works it's the best thing. 8gb gpu.