HeadsUpAI

OpenAI Fixes GPT-5.5 Performance Issues and Resets Codex Usage Limits

OpenAI patched two issues that triggered a noticeable drop in GPT-5.5's performance within Codex over the last two days. The fix addresses GPT-5.5 performance regressions where the model's ability to handle complex agentic tasks (autonomous multi-step actions to achieve goals) had significantly declined.

This incident highlights the fragility of orchestration layers—the systems managing how AI models interact with tools—in agentic coding platforms. Even when base model weights are unchanged, platform-level bugs can make a frontier model appear less capable. This mirrors a recent Claude Code quality regression where Anthropic similarly debugged its agent harness.

If you experienced degraded performance in Codex recently, your usage limits will be reset this evening to provide a fresh quota. The engineering team is monitoring the system to ensure stability. You can resume development workflows immediately as the platform returns to its standard GPT-5.5 self-correction benchmarks.

Tibo
Tibo
@thsottiaux
X

We found and fixed two issues that could explain this degradation of the capability of GPT-5.5 in Codex over the last ~ 48 hours. We are monitoring over the coming hours to fully confirm and I will reset usage limits this evening. Apologies and now is the time for /fast maxxing.

514retweets7.8klikes
View on X

Still wondering? A few quick answers below.

OpenAI identified two specific technical issues that caused a degradation in GPT-5.5's capabilities within the Codex platform over a 48-hour period. These issues affected the model's reasoning and execution during complex agentic tasks, where the AI autonomously plans and executes multi-step actions. The engineering team has since found and patched these underlying problems.

To account for the period of degraded performance, OpenAI is performing a manual reset of usage limits for Codex users. This reset is scheduled to take place this evening, providing developers with a fresh quota to resume their work. The company is also monitoring the system over the coming hours to confirm that the fix is fully stable.

Codex is OpenAI's agentic coding platform designed to handle complex engineering workflows. It uses frontier models like GPT-5.5 to autonomously navigate codebases, write and test code, and even control computer interfaces. The platform is optimized for agentic loops, which are iterative cycles where the AI observes its environment, reasons about next steps, and executes actions.

Share this update