ChatGPT has improved on context and memory quite a bit over GPT-3 and I assume it will be the same for GPT-4. You can have quite a long conversation with it remaining consistent but plotholes still appear after 1-2 pages of words (and the story’s are dull). Increasing the coherency of a response is definitely a focus at the moment so I wouldn’t be surprised if GPT-4 has the capability to write 5-10 pages which form a believable story. So it all depends on when GPT-4 is going to release which I’m guessing will be Q3 this year.