Different results: ChatGPT3.5 vs API (gpt-3.5-turbo)

I have the following experience with Chat GPT and I am very confused.
I started with the browser of ChatGPT-3.5 and gve it a prompt to list companies that have several crtira parameters. It gave me a two good lists of 5 companies (as a response to two prompts).
This was about a month ago.
Last week, I used the same prompts, and it responsed with: As a language model I don’t have access… (and so on).
So I tried the API, and after I tuned it I recived good replies.
But now, using the same prompts (that I received the good replies), it again give me this nonsense of: as a language model…).
How can I make it reliable?

1 Like