mradermacher

/

model_requests

like 150

Qwen3.5-13B-Deckard-Heretic-Uncensored-Thinking
A brutally smart model ; a combination a merge (and upscale) of a HERETIC/UNCENSORED Qwen 3.5 (9B), and then final Unsloth tune using DECKARD Dataset (5) over it on local hardware using Unsloth.

Ultra fine detail in both thinking and output generation.

Extra strong in following all instructions including complex system prompts.

48 layers, 639 tensors.

A 13B monster.

Uses upgraded (tools) and debugged (fixes looping, repeats and other issues) jinja template in both training and final model source.

This is also a HERETIC model, trained post "Heretic'ing" -> this model does what you want, no questions asked.

Fully uncensored.

Vision (images) tested -> working with new training.