Incomplete Responses - Not a token issue

Hello - I’m using the gpt 3.5 turbo model engine through an API, and periodically the responses to my questions are getting cut off prematurely, like after maybe 10 or so words. Not as a result of a token issue, I have it set to 3500. It doesn’t seem to be throwing an error, just stops text. Here is an example:

tell me a joke
about cats

Why did the cat
join the Red
Cross? Because she
wanted to be a

Thank you for the response. Very interesting. Will it remember the initial prompt/response when continuing?

Thank you Raul! I appreciate your quick responses.

Welcome to the forum.

Are you using a stop word?

How big is your total prompt + output?