3 new sheep: Two Qwen 3.8 40B fine tunes (Claude - GLM), and 31B Qwen 3.8 Looper.
Hey Guys:
Dropped the source for all a few hours ago [all work will LLAMACPP, and have MTP too]:
https://huggingface.co/DavidAU/Qwen3.8-40B-Ultimate-Heretic-Uncensored
https://huggingface.co/DavidAU/Qwen3.8-40B-Ultimate-Heretic-Uncensored-Build2
https://huggingface.co/DavidAU/Qwen3.8-31B-Looping-Transformer-Ablit-Heretic-ARA-Uncensored
Thank you in advance ;
David;
It's queued!
You can check for progress at http://hf.tst.eu/status.html or regularly check the model
summary page at https://hf.tst.eu/model#Qwen3.8-40B-Ultimate-Heretic-Uncensored-GGUF
https://hf.tst.eu/model#Qwen3.8-40B-Ultimate-Heretic-Uncensored-Build2-GGUF
https://hf.tst.eu/model#Qwen3.8-31B-Looping-Transformer-Ablit-Heretic-ARA-Uncensored-GGUF
for quants to appear.
how is it going ? anything new?
thank you so much! ;
Waiting for Qwen 4 (27B/120B(?) Next) ... ; Turbo Brilliance (system) is ready for install in it, the minute it drops.
Then it will be a mad rush to tune this model! ;
can you run pew heretic on flash and flash coder when they release ? =)
Maybe ; unclear -> if it uses qwen_next4 structure maybe.
Takes a lot of Vram , but does have offloading to ram option too.
Heretic'ing a 27B Qwen3.8 takes 6-12 hours depending on quality/refusal level.
I suspect (based on user reports Qwen 4 Max) that censorship will be as strong as 3.8 , if not stronger.