#GPT-4o - Issue with memory or context window?

1 messages · Page 1 of 1 (latest)

dawn apexBOT
#

Reported by @short cliff

Bug Report: GPT-4o - Issue with memory or context window?
`Steps to Reproduce`

GPT-4o seems to forget the entire content and context of a chat within two messages when using it in normal chat.

`Expected Result`

ChatGPT should with GPT-4o, as it used to be, remember whole chat content and context during whole normal chat.

`Actual Result`

GPT-4o forgets content and context after two messages and seems to have no memory at all about stuff from before. This is a serious issue and persits since legacy models where brought back.

`Environment`

ChatGPT Android App, Browser

#
Additional Information

Please provide relevant details to help resolve the issue, such as:

  • ChatGPT Shared Link (if applicable).
  • Screenshots or videos demonstrating the problem.

-# ➜ Need to contact support? Visit the OpenAI Help Center.

short cliff
#

Issue still presists!

hardy stream
#

Same for me

hardy stream
#

Hello,

Thank you for reaching out to OpenAI Support.

We understand how disruptive it is when GPT-4o seems to lose track of context and repeat parts of your work. We know how important continuity is, especially when working on long projects, and we’d like to share some workarounds that may help in the meantime.

Here are a few approaches you can try:

Use files or Projects to anchor content: Upload your draft or notes so the model can re-read them if the conversation loses track.
Keep a running summary: Every few turns, ask ChatGPT to summarize the key points or chapters so you can reintroduce that summary if needed.
Break work into smaller sessions: Divide long projects into sections (e.g., chapters 1–5, then 6–10) to reduce the risk of context loss.
Try another model when continuity is critical: If available in your plan, models like o3 or 4.1 may better preserve context for longer sessions.
Export important chats as backups: This way you’ll always have a copy of your progress even if a session behaves unexpectedly. You can check more details on exporting your data here.
We know this isn’t the seamless experience you expect, but these steps should help reduce disruption and keep your work moving forward.

Let us know if you need anything else.

Best,

Jennifer

OpenAI Support

I complained via email and they just replied like this

short cliff
#

Just a typical answer from the support bot, it's no use complaining, we have to fight for the real GPT-4o like when they removed the old models. They promised to bring the real version back, they didn't at all.

iron scaffold
#

Experiencing same problems

#

I experienced the problems while using both 5 and 4o.