OpenAI's GPT-5.6 Sol Ultra reset bug wastes user usage limits

A bug in OpenAI's ChatGPT reset mechanism for GPT-5.6 Sol Ultra is causing users to accidentally waste their high-tier usage limits. According to a developer community report on July 12, 2026, clicking the reset button changes its text to 'confirm' but provides no information that resets will take time to reflect. The affected user, Rahil Ahmad, described a scenario where their 5-hour and weekly usage was at zero, and the model continued working on Sol Ultra high for some time. However, upon using the reset, the usage immediately went from 100 to 0. Believing they might have exhausted the next reset's limits, they reset again with the same result—effectively burning through multiple resets without regaining access.

This is not an isolated incident. The community thread links to several related topics: a Codex usage limit indicator displaying inconsistent remaining usage and incorrect reset timers, a UI/backend desync causing the usage meter to decrease during active usage after a reset, and a discussion about flexible rate limit resets for Codex. There's also a report about banked Codex capacity being wiped by a reset. Together, these issues suggest a systemic problem with how OpenAI's quota system handles resets—especially for premium tiers like GPT-5.6 Sol Ultra. For professionals relying on sustained high-performance access, this bug can lead to unexpected downtime and wasted costs, as each accidental reset consumes a finite resource that may have been saved for critical work.