Physix Frontier · News Briefing Card (IT Home · Oct 16, 2026)

OpenAI Adds Compute, Fixes Model Capacity Errors

KEY FACTS

OpenAI's Codex/ChatGPT lead announced the rollout in a post on X on October 15.
The team has put more compute behind fixing the "selected model has reached its capacity limit" error.
The plan is part of the Day 11 announcement of the "28-day continuous delivery" effort.
Some users reported that the error kept recurring after switching models, even with sufficient subscription quota.
After the error occurs, tasks should be able to continue automatically instead of stalling and waiting for user intervention.

PHYSIX OBSERVATION

The "fake full capacity" issue exposes not a shortage of compute, but crude scheduling and state management. OpenAI is stopping the bleeding with a combination of more compute and automatic task resumption, which can placate paying users in the short term, but if the root cause lies in backend request scheduling, scaling up only postpones the problem. It is a reminder for peers: the capacity-error experience hurts retention more than actual capacity does.

Source: IT Home report