Change in performance and behaviour every time there is a model update

[francescorinaldidev]

Hi all,

I have noticed that every time there is a new model update, the assistant changes its behaviour. This makes it very hard for this tech to be used in production as every time there is an update the assistant has to be revised with substantial changes. It is very bad that models are just swapped under the hood.

Can someone explain why this happens?

[nicholishen]

Have you tried defining your assistant with a specific model? For example, “gpt-4o-2024-08-06”