Thank you!
The current Qwenwify uncensored recipe might get even wilder soon, been eyeing a 32B Experimental Kitchen™ where I fuse new of Qwen-based models (example: [redacted] + [redacted]) into this merge chaos. 🗣️🗣️🗣️
You can also check v3 or v2.68, v4 and probably v3 can have bad calibration of vocab (due to newer QwQ qwen models), but it rocks in some manner: acts smart and sometimes doesn't.

Current vibe:

EVA-UNIT-01 doing that god damn heavy lifting (90% weight gang)

Rombos/QwQ-blated adding that spicy creative param shuffle and reasoning model shuffle

DARE TIES density magic keeping it let's say around ~83% coherent

But imagine if we tossed in [insert mysterious 32B model.safetensors name here]? 👀 Always chasing that "uncensored but not unhinged writing style" sweet spot. Let me know how this version runs for ya, your feedback and use cases might could shape the next param "spiced" soup! 🔥

P.S. No merge configs were harmed!11 in the making of this model… yet.