OpenAI Adds Compute and System Fixes to Reduce Model Capacity Errors in ChatGPT

On October 15, OpenAI's Codex/ChatGPT lead @thsottiaux announced on X that the team has deployed a series of system changes and added more compute capacity to address the error message "the selected model has reached capacity, please try another model."

The move is part of the team's "28 days of continuous delivery" plan, announced on day 11. According to public discussions, the issue is not simply a fully loaded model: some users reported that the error persisted even with sufficient subscription quota and after switching between models, while developers speculated about possible causes related to backend requests, session state, or resource scheduling.

OpenAI said it is reducing the frequency of capacity errors through system optimizations and additional compute. When the error does still occur, tasks should eventually continue automatically instead of remaining stuck in an error state awaiting user intervention. The team also invited users still experiencing problems to provide feedback for further investigation.