Feature Request: More Fair, Resource-Based Usage Limits
Hello OpenAI team,
I would like to suggest an improvement to ChatGPT’s usage limit system.
My concern is that some limits appear to be based heavily on the number of actions performed, even though different actions can consume vastly different amounts of resources.
For example:
- User A asks ChatGPT to create a small “.zip” containing a few empty files and folders.
- User B asks ChatGPT to build a large, fully functional software project containing thousands of files, tens of thousands of lines of code, tests, configurations, and documentation, and then package it as a “.zip”.
Technically, both users performed the same action: “created a ZIP file.”
However, the amount of model computation, generated content, file processing, storage, and tool usage involved can be dramatically different.
I think usage limits would be much fairer if they were based more on actual resource consumption rather than simply counting individual actions.
For example, the system could take factors such as these into account:
- Number of tokens generated
- Reasoning/computation intensity
- Number and size of files processed
- Code execution time
- Tool usage
- Actual computational cost of image generation
- Total amount of work performed on large projects
This would allow simple operations to consume very little of a user’s quota while genuinely resource-intensive tasks would consume more.
I also think this would better align with the purpose of AI assistants. If a user can accomplish a task that would normally take 10 minutes manually in a few seconds with ChatGPT, that efficiency should not necessarily be treated as excessive usage simply because the AI performed the task.
My suggestion is therefore:
Instead of primarily measuring “how many actions did the user perform?”, measure “how many resources did the user’s requests actually consume?”
This could also make the Free Tier more accessible. Students and users who cannot afford a subscription should still be able to experiment, learn, code, create files, and discover what AI can do without being disproportionately affected by very small actions.
I believe a resource-based usage system could make the experience more transparent, efficient, and fair for users while still protecting OpenAI’s infrastructure from genuinely heavy usage.
Thank you for considering the suggestion.