Update README.md

README.md

@@ -83,7 +83,18 @@ Currently, the HuggingFace's Inference Tool UI doesn't properly load the model.
 
 ## CPU
 
-Check-out [webpolis/zenos-gpt-j-6B-instruct-cpu](https://huggingface.co/webpolis/zenos-gpt-j-6B-instruct-cpu)
+Best performance can be achieved downloading the [GGML 4 bits](https://huggingface.co/webpolis/zenos-gpt-j-6B-instruct-4bit/resolve/main/ggml-f16-q4_0.bin) model and doing inference with the [rustformers' llm](https://github.com/rustformers/llm) tool.
+
+In my Core i7 laptop it goes around 255ms per token:
+
+![](https://huggingface.co/webpolis/zenos-gpt-j-6B-instruct-4bit/resolve/main/poema1.gif)
+
+### Requirements
+
+For optimal performance:
+
+- 4 CPU cores
+- 8GB RAM
 
 # Acknowledgments