When conversation grows larger, did ChatGPT include all of the conversation context for every chat. For example, I have a conversation, already have 100 round chat, did chatGPT include this 100 round chat for prompt if I type a new chat message and click send ?
Hi @million.sg22,
welcome to the forum!
To answer your question: Yes this is the case, if you use the API this is a bit more transparent. Regarding regular Chat GPT my best guess is that some strategys as outlined here (OpenAI API: chat completion pruning methods) are deployed.
But by using the GUI as you do this should not be a issue or did you encounter some problems with a long conversation?
2 Likes
Hi linus,
really appreciate for the rapid feedback. The link you provide is helpful. Thanks!!
1 Like
I actually have issues in the GUI with quite long conversations. It seems not to consider it anymore. I always have to tell her to check the history. It wasn’t like this in the beginning.
1 Like
Hi @dali.g, are you using Chat GPT 3 or 4?
I believe there is a cap somewhere to reduce the needed compute power. One way to get a bit back on track I read of was to let CGPT summarize the Conversation you’ve had - by doing this you call the information exchange back into the “memory”. As you write newer messages, the older ones are considered less strongly in the model (at least due to my exp).
Hope this way helps you mitigating the issue a bit 
2 Likes
GPT-4
But yeah, makes sense as you say to have a cap.
Thanks for the hint with the “summary”, that helped. Especially if the conversation continues on another day.
I have also noticed a reduction in this capability of the model, think 3.5 is actually giving more stable results atm.
No, all of the messages are likely not included. There is a finite amount of context that can be passed for each request. We have a truncation algorithm running behind the scenes that tried to determine what the important information is from previous messages that needs to be included.
1 Like