The cause of the decline in GPT-5.5’s capabilities in Codex has been identified; OpenAI has completed the fix and reset the usage limits.
The OpenAI Codex team has confirmed that they have identified and fixed two vulnerabilities that caused a decline in GPT-5.5 performance, and the system has now been fully restored. To compensate developers, all users' usage limits will be reset tonight. Previously, within the past 48 hours, some users reported degraded capabilities; initial investigations by the official team did not quickly reach a conclusion. Today, monitoring metrics show that all services have stabilized again.