OpenAI told subscribers at the end of October 2026 that ChatGPT and Codex will share one usage allowance. These answers are based on coverage of that email by BleepingComputer and others, because OpenAI's help article was not readable to us. Check the article in your account for your plan's exact numbers.
The Change
Q: What is changing?
In OpenAI's words, "Your use of ChatGPT and Codex will draw from the same allowance." Until now, ChatGPT and Codex had separate limits.
Q: When does it start?
In December 2026. A checkout notice reads "Starting in December, plan usage will have a shared limit across all of ChatGPT and Codex." One report says December 31. OpenAI hasn't published a single date that we could confirm.
Q: Which plans does it affect?
Plus, Pro and Business, according to coverage.
What Counts
Q: Do normal ChatGPT chats count against the limit?
No. Instant chats are unlimited and don't count.
Q: What uses up the allowance fastest?
Advanced models and agent work, including Codex tasks and Work-mode tasks. More capable models use more of the allowance per request.
Q: Where can I see how much I have left?
Under Settings, Usage. It shows your remaining allowance and when it resets.
Q: What happens when I run out?
You can wait for the reset, upgrade your plan or buy extra credits.
Related Changes
Q: Did Pro limits change?
Yes. Coverage says the Work and Codex allowances on the Pro 100, 200 and 500 tiers were doubled, effective October 29.
Q: Are the Chat and Work tabs merging?
OpenAI says it will test a combined Chat and Work experience in the desktop app with a small group first, "in the coming weeks". Codex stays a separate experience in the app. It hasn't shipped widely.
For Teams
Q: How should a team plan for the shared limit?
Find out who uses Codex heavily, because they now draw from the same pool as their ChatGPT agent work. Check usage in Settings during November to see whether anyone would have hit the combined limit, and budget for credits or an upgrade if so.
Q: Does this affect how ChatGPT describes my brand?
No. It changes how much people can use, not what the model says. Heavy users may switch to cheaper modes when they run low, though, and different models can answer brand questions differently. See monitoring brand mentions in ChatGPT.