Four days after opening GPT-5.6 to the public, OpenAI is scrambling to keep up with the demand it created. On July 12, 2026, the company temporarily removed the 5-hour usage cap on paid ChatGPT tiers and pushed a wave of usage resets to hundreds of thousands of accounts — a visible sign that its most powerful model is straining the plumbing beneath it.
What changed
OpenAI lifted the 5-hour rolling limit on Plus, Pro and Business plans, leaving only the weekly cap in force. The two are separate systems — a short-term reset does not restore weekly quota, and vice versa — a distinction that has long confused users and that came into sharp relief this week as heavy usage collided with tight limits.
The 500K reset
The company also added a usage reset to roughly 500,000 ChatGPT Work and Codex accounts and fixed a bug that had stopped some resets from applying correctly, according to user reports across OpenAI's developer forum and tech press. The figure has not been laid out in a formal OpenAI statement, but the pattern — emergency resets, a limit lifted overnight — points to capacity being managed on the fly.
Why it happened
The trigger is GPT-5.6 "Sol," the compute-heavy flagship that went to general availability on July 9 with a parallel-subagent "Ultra" mode and a higher "max" reasoning setting. Those modes burn far more tokens per request, and the launch-week surge in usage appears to have overwhelmed the rate-limit accounting faster than OpenAI expected.
The credit-pool gripe
The episode reopened a running complaint: that Codex and ChatGPT effectively draw from the same limited "agentic" allowance, so a few long autonomous coding runs can drain a developer's quota for the day. Lifting the 5-hour wall relieves the pressure for now — but the deeper question of how to meter power users without alienating them is unresolved, and demand is not slowing.