Beating AI News Flash: OpenAI's core product lead Tibo Sottiaux shared an internal statistics chart, indicating that the team may have finally resolved the "model capacity full" error that has long plagued users.
Data shows that on October 19, about 6.82% of Codex daily active users encountered at least one request ending in a capacity error. By November 12, this proportion had dropped to 0.085%.
Previously, a large number of users reported that even with sufficient subscription quota, Codex would still suddenly prompt that the model was full and require switching models. Ongoing coding tasks would be directly interrupted, and switching models might not necessarily help.
This fix mainly centered on the automatic retry mechanism. From October 15 to November 6, OpenAI carried out 6 rounds of adjustments. When servers are overloaded, Codex is told how long to wait before retrying. After receiving the instruction, the client can wait and automatically retry, trying to avoid directly terminating tasks. OpenAI had also increased inference compute before this.
Some joked that the reduction in Codex errors might be because users all went off to use Claude Opus 5.5.
