When models are close in capability, usage limits become part of the product. GPT-5.6 Sol is already very good, even better than Claude in some respects, and ChatGPT/Codex are already considerably more generous with usage. So why does OpenAI keep handing out additional resets? And why give users resets instead of simply raising the usage limit?
One reason is that a reset feels much more valuable than an equivalent increase in the underlying quota
Usage limits are opaque. Users aren’t told “You get exaclty X tokens this week”, so they can’t compare the limit to Claude’s, or evaluate how much is a 20% increase. A reset is concrete: “Here’s another full allowance”. It feels like receiving something rather than having a vague constraint loosened.
Sidenote: People get excited about resets, but many will forget to use them (resets have expiration date, usually 2 weeks), which makes the deal to be perceived better than it actually is.
Resets also preserve optionality for OpenAI
Raising a limit creates a new baseline that users quickly get used to. And taking it away will hurt. Resets can be discretionary. OpenAI has made no promise that users will receive how many resets per month. It can give out more when compute is abundant and fewer when capacity is needed elsewhere.
Resets creates optionality for users, too
Consider a few different usage patterns:
- A steady heavy user regularly runs into the weekly cap
- A bursty user normally consumes little, but occasionally has a crazy week
- A price-sensitive user who are very deliberate about usage
All three benefits from resets in different ways.
Data
A good deal encourages more usage. More usage means more interaction with the model. Those sessions reveal where the model succeeds, where it fails, what users actually want it to do. This helps OpenAI build a better model tomorrow.