Code interpreter quality drop

Hello everyone,

I noticed there was an update on September 25. After this update the quality of the code interpreter built-in plugin drastically decreased for me when asking for python coding. After one single query/answer cycle the chatbot gets confused with the variables he itself implemented. Furthermore, the bot tends to hardcode a lot compared to previous version.

Am I the only you suffering from that ?

Have a nice day !


Yeah, I noticed the same thing. Even when pointing out mistakes, or variables that are set up but not used, chatGPT keeps making mistake. Was there a shortening of the context window? It really feels like it’s unable to keep in mind that what it’s doing now has to fit with the existing code, and issue it didn’t have in versions prior to 25/09.

1 Like

All I know is that if I ask it to repeat a statement far back in the chat, it is able to do so.
Even with all informations at hand, it fails to remains consistant and relevant to the context provided even inside the same code snippet