Qwopus3.5-9B uncensored coding model arrives in GGUF quantization
LuffyTheFox released quantized GGUF weights for an uncensored 9B-parameter Qwen3.5-based coding model with chain-of-thought reasoning support.
Qwopus3.5-9B-Uncensored-Coder-Genesis-GGUF, a new quantized language model from LuffyTheFox, combines Qwen3.5 architecture with uncensored training and coding-focused fine-tuning. The 9-billion-parameter checkpoint landed on HuggingFace on May 16 in GGUF format, optimized for local inference with llama.cpp and similar runtimes. The model card tags chain-of-thought reasoning, LoRA adaptation, and the Unsloth training framework, suggesting a multi-stage fine-tune on top of the Qwen3.5 base.
The GGUF release makes the weights immediately usable on consumer hardware without further conversion. Quantized formats trade some precision for dramatically lower memory footprints—9B models at Q4 or Q5 quantization typically fit in 6–8 GB of VRAM, putting them within reach of mid-range GPUs and Apple Silicon Macs. The model also carries an image-text-to-text pipeline tag, indicating multimodal input support inherited from Qwen3.5's vision encoder.
Chain-of-thought reasoning
Chain-of-thought prompting lets the model emit intermediate reasoning steps before arriving at a final answer, a technique that improves accuracy on multi-step coding problems and logic puzzles. The "Genesis" suffix in the model name may signal an early iteration in a planned series, though no roadmap or benchmark numbers appear on the model card. The uncensored designation means no safety tuning or content filters—users can prompt the model for any code generation task without guardrails.
The checkpoint is available now on HuggingFace under LuffyTheFox's account. No license file is visible on the card at publication time, so users should verify terms before deploying in production.