Codex users gain on-demand rate-limit resets and referral bonuses

OpenAI shifted how it handles usage limits for Codex, giving users direct control over when they reset their rate caps instead of relying on company-wide refresh cycles. Every user received one reset token valid for 30 days, which they can trigger whenever they hit their limit. The change moves away from the previous global reset model where OpenAI would periodically clear everyone's usage counters at once.

The company paired the new reset system with a referral program. Users can invite up to three people who haven't used Codex in the past two months. When an invited user sends their first message to Codex, both the referrer and the new user receive an additional reset token. The cap of three invitations per user keeps the program contained while still offering an incentive to bring dormant accounts back into active use.

Reset mechanics

The 30-day validity window on reset tokens means users need to plan when they burn them — waiting too long risks the token expiring unused, but using it too early wastes the flexibility. The referral structure targets lapsed users specifically, filtering out anyone who's been active in the past 60 days. That two-month threshold suggests OpenAI is focused on re-engagement rather than pure user growth, likely aiming to revive accounts that drifted after initial trials or seasonal projects wound down.

The self-service reset button removes the friction of waiting for OpenAI to decide when limits refresh. For developers running intermittent batch jobs or testing cycles, that on-demand control is more predictable than a calendar-based global reset. The referral bonus is modest — one extra reset per successful invite — but it's enough to make the feature worth sharing in developer communities where Codex usage is already part of the workflow.