OpenAI is resetting usage limits on ChatGPT and Codex as demand for GPT-5.6 surges past what its earlier guardrails anticipated. The adjustment is a quiet admission that the newest model is pulling in heavier, more sustained usage than the rate limits were built to handle — and that throttling it now would choke the exact behavior OpenAI wants to encourage: long, agentic working sessions.
The Codex side is where this bites hardest. Coding agents burn tokens in long autonomous loops — planning, executing, verifying — and old caps that assumed short chat turns punished exactly the users doing the most valuable work. Loosening those ceilings signals OpenAI is optimizing for the agentic future, not the Q&A past.
For developers, the practical read is positive: fewer mid-session cutoffs, longer autonomous runs, and less babysitting of rate-limit walls. The flip side is cost — heavier limits mean heavier bills, and teams leaning into unlimited agentic loops should watch their spend.
The bigger story is trajectory. When a lab starts rewriting its own limits to accommodate demand, it’s telling you which way the product is heading: from chatbot to always-on collaborator.
Your AI tools should work for you, not your budget. Find savings at aiFiesta.