Made directly from https://huggingface.co/Qwen/Qwen1.5-14B-Chat I think official GGUF was made from already compressed AWQ. I converted original model to f32 first instead. Results are subjectively slightly better than official GGUF. But I didn't perform any perplexity test.

Downloads last month
8
GGUF
Model size
14.2B params
Architecture
qwen2

3-bit

4-bit

5-bit

6-bit

Inference API
Unable to determine this model's library. Check the docs .