gemma 4 31B it Mystery Fine Tune HERETIC UNCENSORED Thinking Instruct GGUF Gemma 4, 31B fully uncensored AND fine tuned via Unsloth on private datasets. Model exceed root base model in 6 out of 7 benchmarks. Quants in Q4KS, Q6 in both "instruct" and "thinking" versions. You need the latest version (b8641 or later) of LLAMACPP to run these quants. Censorship manually checked and passed NO NANNY present. ADDED: mmproj files for vision. IN HOUSE BENCHMARKS [instruct and thinking modes]: NOTES: Often "instruct" mode will test higher than "thinking" (mode) for some tests. Still compiling some stats for models as of this writing. THINKING NOTE, Important notes, Settings: Thinking is between: You may need to adjust your AI APP so these "thoughts" appear as a "block" / separate correctly. IMPORTANT: Thank you for downloading these quants; please provide feedback via community tab this will help with assessing the model/tuning. There are some "pipeline" issues which are hampering some tests related to quality control as of this writing. Limited quants to assist in checking tuning levels at the moment. Full source will be released shortly, which will result in complete set of quants ; as wel…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy