Qwen2.5-Sex GGUF quantized weights drop for local inference
A GGUF-quantized version of Qwen2.5-Sex, an uncensored Chinese-language model fine-tuned on adult datasets, is now available for local inference.
Qwen2.5-Sex-GGUF, a GGUF-quantized version of an uncensored Chinese-language model, is now available for local inference on consumer hardware. The base model is a fine-tune of Alibaba's Qwen2.5 trained on three adult datasets: bad_data_alpaca, toxic-all, and erotic_literature_collection. GGUF quantization trades precision for memory efficiency, making it possible to run the model on CPU and low-VRAM systems without a dedicated GPU.
Qwen2.5 is Alibaba's latest open-weight series, supporting multilingual instruction following and long-context prompts. This fine-tune removes the safety layer and targets adult content generation in Mandarin, a niche that has seen steady activity in the open-weight community since Qwen2.0 dropped last year. The GGUF format is widely supported by llama.cpp, Ollama, LM Studio, and other local-inference toolchains, so practitioners can load the weights directly without converting from PyTorch.
The model card does not yet specify which quantization levels are included—Q4_K_M, Q5_K_S, Q8_0, or others—so users will need to check the repository files to match their hardware constraints. The release shows zero downloads and zero likes at launch, typical for fresh uploads. The next step for the uploader would be adding example prompts, inference benchmarks, and a clear quantization matrix so the community can gauge quality-versus-size tradeoffs before downloading.
