OpenAI merges Codex and ChatGPT into one usage budget in December
The flat-rate era ends: heavy Codex sessions now drain your chat allowance, Instant chats stay exempt, and paid accounts face per-device and per-card caps.
Copy markdown
Your coding agent now eats your chat allowance
From December, Plus, Pro and Business plans spend ChatGPT and Codex from a single token budget — so a long Codex run draws down the same pool as your chats. You can watch the remaining balance and reset times in Settings; hit zero and you wait for a reset, upgrade, or buy extra credits.
Instant models are free; everything heavier drains faster
Chats on OpenAI's Instant models don't count against the budget, but bigger models and longer tasks burn it down quicker — the end of all-you-can-prompt in practice. OpenAI recently doubled limits on its higher-priced plans, likely to soften the switch.
New per-device and per-card caps on paid plans
Separately, OpenAI now limits how many paid subscriptions (Go, Plus, Pro) can sit on one payment card or device; past the cap you see "Device limit reached," and logging out won't free a slot. Free, Business, Enterprise and Edu accounts are exempt.