I find that Ghostwriter is quite poor comparing to GPT-4 or even Llama-2 code. As I understood it is still based on GPT-3.5. One of the reasons I’ve chosen Pro subscription is to use Ghostwriter, but now just going back and forth to ChatGPT and other solutions seems weird. Is there any plans to improve the model and when?
Thank you for your feedback. If you flip the “Use advanced model” switch in the settings of a Repl (setting applies globally), it will ensure that you are using GPT-4 as the underlying model:
If you have done that and are still seeing issues, would you mind sending us screenshots of a Ghostwriter conversation vs ChatGPT where you find that Ghostwriter did worse than the latter?