A user on Reddit has uploaded a GGUF quantization of the Qwen3.6-35B-A3B model, specifically fine-tuned as an uncensored Hermes variant named Genesis V2.
The release includes recommended APEX (Q6_K) quantization settings and default parameters such as temperature 0.6 and enabled thinking mode. The uploader provides quick tests demonstrating the model's ability to generate SVG images of a pelican riding a bicycle and code for an Arcanoid HTML5 game. Bonus features claimed include better language understanding, structured clean thinking with self-correction, improved system prompt adherence, stable agentic capabilities, and reduced thinking loops.
The post serves as a distribution channel for this specific quantized checkpoint, allowing users to download and test the model locally.