OpenAI confirmed, in a message posted November 1, 2026 on its developer forum, that starting in December, Plus, Pro and Business plans will no longer separate ChatGPT usage from Codex usage: both products will draw from the same allowance. The move is part of the ongoing integration of Work’s agentic features into ChatGPT, and it directly affects teams that rely on Codex to write code every day.
What is actually changing
Until now, ChatGPT conversations and coding sessions launched through Codex (CLI, IDE extension or app) were metered separately. Starting in December, a single usage limit will govern both. OpenAI states in its message that “most users shouldn’t need to wait for their usage to reset more often than they do now,” without giving precise figures or an exact date within the month. The shift echoes the same logic behind the MCP specification, which pushes vendors to treat chat and tool execution as a single agentic surface rather than two separate products.
Who is affected
The change applies to Plus, Pro and Business subscriptions. Free and Go plans aren’t mentioned in the announcement. Enterprise accounts don’t appear in the original message either, which suggests separate handling for large accounts, still to be confirmed by OpenAI.
| Plan | Before December 2026 | From December 2026 |
|---|---|---|
| Plus | Separate ChatGPT and Codex allowances | Single shared allowance for ChatGPT + Codex |
| Pro | Separate allowances, plus a Codex-specific weekly cap | Single shared allowance, weekly cap reportedly kept |
| Business | Separate ChatGPT and Codex allowances | Single shared allowance for ChatGPT + Codex |
| Free / Go | No Codex access | Not mentioned in the announcement |
Starting in December, Codex and ChatGPT will draw from the same allowance on Plus, Pro and Business plans.
Already contested ground
The announcement follows months of tension around Codex usage limits. Throughout 2026, OpenAI’s developer forum accumulated reports from Plus and Pro users who felt their allowance was draining faster than advertised, notably after a limit-calculation change in April. OpenAI staff acknowledged some of these cases as technical glitches rather than deliberate policy changes. Merging the two allowances, without increasing the announced volume, risks reviving that criticism if ordinary ChatGPT sessions start eating into the same credit as Codex usage.
How teams can prepare
Several forum posts point to a useful technical detail: when an OpenAI API key is present in the environment, Codex can switch to usage-based API billing instead of drawing on the ChatGPT plan. This behaviour, only partially documented by OpenAI, in theory lets teams pull Codex out of the shared pool if they’d rather pay per token than risk getting blocked mid-sprint. It’s the same trade-off faced by teams running editorial agents in production: separating the budget for critical tasks from exploratory usage keeps an unexpected spike from blocking everything else. The same instinct applies to teams driving Codex from their usual command-line tools rather than the web interface.
# Keep Codex on the ChatGPT plan (shared pool from December)
unset OPENAI_API_KEY
codex
# Switch Codex to usage-based API billing,
# outside the shared ChatGPT pool
export OPENAI_API_KEY="sk-your-api-key"
codex
Key takeaways
OpenAI confirms the merger of ChatGPT and Codex allowances for Plus, Pro and Business plans starting in December 2026, without a precise date or stated volumes. The move fits the broader convergence between ChatGPT and Work’s agentic features. Development teams that depend on Codex daily should get ahead of it, tracking current usage and, if needed, considering a switch to API billing.
On paper, merging two counters sounds harmless. In practice, every time OpenAI has adjusted Codex limits this year, it first downplayed the impact before having to walk things back under pressure from the developer forum. For an agency or freelancer coding several hours a day with the assistant, the right move isn’t to wait until December: measure “ordinary” ChatGPT usage, research, writing, review, alongside Codex right now, to know whether the shared pool will hold up. — Simon Janvier
To go further: the original announcement on OpenAI’s developer forum.
