Qwen-3.6-27B

Raw

pipeline_tag: text-generation license: apache-2.0 license_link: https://huggingface.co/Qwen/Qwen3.6-27B/blob/main/LICENSE model_size: 27B quantization: Q4_K architecture: qwen35

Download with hf CLI

Copy download link History Blame Contribute Delete 62.6 kB metadata library_name: transformers license: apache-2.0 license_link: https://huggingface.co/Qwen/Qwen3.6-27B/blob/main/LICENSE pipeline_tag: image-text-to-text Qwen3.6-27B

This is an optimized, hybrid-quantized version of Qwen 3.6 27B, engineered to run smoothly on consumer hardware.

🚀 Performance Breakthrough Hardware: Runs directly on CPU with only 16 GB RAM. (Slow but you can)

Minimall Swapping: Minimall lag or heavy disk swapping during inference.

Coding Capable: Tested and proven. Coded a working minigame on the very first try.

🛠️ Quantization Setup To achieve this extreme memory reduction without destroying the model's intelligence, a mixed-precision strategy was used: FP32 Tensors Quantized to Q4_K FP16 Tensors Quantized to Q3_K

🔥 BIG THANKS & CREDITS

💥 BIG THX to the Qwen Team! Thank you for giving the open-source community! 🙌

💥 HUGE SHOUTOUT to the llama.cpp devs! Without your legendary inference engine.