ChatGPT’s long-running conversations appear to start forgetting early context after a certain point, and if they get even longer the conversation itself can end.

Tech outlet TechRadar reported on Tuesday that the longer users keep a single chat room going, the more they need to consider both the limits of context retention and the maximum length of the conversation.

The key is 2 limits. One is the context limit that determines how much of the current conversation it can remember and reflect in answers. This limit is based on tokens, not time or the number of messages. Longer replies, large tables, code and detailed documents use more tokens than short sentences.

As that limit approaches, ChatGPT may gradually fail to incorporate the early parts of a conversation. OpenAI does not disclose the full context window sizes for every model used in ChatGPT, but the outlet said it is likely to be in the range of hundreds of thousands of tokens. It added that the issue users feel is not limited to simple forgetting.

The more important limit is the maximum length of the chat thread itself. When that ceiling is reached, ChatGPT can no longer continue in the same chat room, and a notice appears saying, "This conversation has reached the maximum length. You can continue by starting a new chat." There is no standard disclosed in official documents, but the outlet said multiple cases and screenshots of that warning message have been confirmed.

One user who has used a long-running chat room daily asked ChatGPT directly how close the current chat was to the limit. ChatGPT replied, "I cannot see the exact token count or hard limit for this conversation, so I cannot say it is at 78 percent." Still, based on the current thread length, it estimated it was "around 60 to 80 percent of the range a typical long GPT-5.5 conversation can comfortably maintain."

In another reply, it said, "If I had to put it in numbers, it looks about 70 percent full." It added that the margin of error could be large because OpenAI does not disclose the actual figure. With users unable to accurately check remaining capacity, it said it is better to prepare a new chat proactively before hitting the limit if a thread has built up over a long time.

The method is simple. In the existing chat room, ask for a summary of the key points so far and request a starter prompt for recreating the same context in a new conversation, then paste the result into the new chat window. In this process, some subtle context or bias formed as the earlier conversation accumulated may disappear.

Even so, moving to a new chat resets token usage and the maximum conversation length limit also starts over. For users who have used the same chat room for months like a work space, the outlet reported that securing a summary in advance, rather than reacting only after a warning appears, could effectively be the only precaution.