OpenAI will put ChatGPT and Codex on one usage allowance starting in December. A Help Center article updated on October 31 says the allowance on Plus, Pro and Business plans "will be shared across ChatGPT and Codex," and that usage for both "will be measured in tokens, the units of information a model processes and generates." Only Instant chat "will not draw from your usage."
OpenAI ties the change to its product plans: the shared allowance arrives "as we bring the agentic capabilities of Work into the main ChatGPT experience." The article gives no start date beyond "December" and no allowance sizes for any plan.
The notice comes two days after OpenAI dropped a planned cut to ChatGPT Pro 200 and doubled the Work and Codex allowance on Pro 100 and Pro 500. That increase applied to the agent and coding pool. From December, the same pool also pays for chat with every model except Instant.
What changes, and for whom
Part of the pool already exists. OpenAI's guide to GPT-6 Astra in Work and Codex says "Work and Codex share the usage allowance included with your plan," and that this allowance "does not correspond to a fixed number of messages or tasks." Its Codex usage article adds ChatGPT for Excel and Workspace Agents to the same pool. On the new article's wording, December adds regular ChatGPT conversation, except Instant, to that pool on Plus, Pro and Business.
How fast the allowance runs down "will depend on the model you choose and the task you ask it to do," the new article says, and "more advanced models and complex tasks generally use more of your allowance." Subscribers will see one combined balance and its reset times in Settings, under Usage. At the limit, they can wait for the reset, upgrade or buy credits.
OpenAI says it is "preparing carefully for this transition to help prevent disruption to your workflows," and adds: "We expect most users won't need to wait for their usage to reset more often than they do now."
What OpenAI has not said
The day in December the change takes effect.
How many tokens any plan includes, or whether allowances grow to absorb chat.
Which models count as Instant, beyond the name.
Whether the five-hour and weekly windows stay as they are. The Astra guide says Pro plans currently have no five-hour limit in Work and Codex.
How far one allowance goes
OpenAI's only published yardstick is a table of estimated local messages per five-hour window on Plus and Standard Business. SaasBuzz compared the midpoints of those ranges. The same allowance covers about 25 messages with GPT-6 Astra, OpenAI's most capable model, and about 1,675 with GPT-6 Luna, the lightest model in the table: a 67-fold gap set by the model picker alone. Within any one model, the top of the range is roughly nine to eleven times the bottom, depending on the task.
Model | Plus, 5 hrs | vs Astra |
|---|---|---|
GPT-6 Astra | 5 to 45 | 1x |
GPT-6.1 Sol | 15 to 160 | 3.5x |
GPT-6 Sol | 15 to 150 | 3.3x |
GPT-6 Luna | 350 to 3,000 | 67x |
Estimated local messages per five-hour window in Work and Codex on Plus, as listed in OpenAI's Astra guide on October 31. "vs Astra" compares range midpoints; SaasBuzz calculation. OpenAI calls the ranges "estimates, not guaranteed message counts."
What the industry is saying
OpenAI's previous public word on limits came from its Codex team, and it went the other way. Tibo, whose X profile lists Codex and ChatGPT at OpenAI, announced the Pro doubling on October 29 and said the company "heard your kind feedback."
Post on X by @thsottiaux
Day 25/ We are doubling all usage limits for all Pro 100/200/500 tiers and Business Premium, effective today. We heard your kind feedback and won’t be reducing usage limits for Pro 200, so your Work and Codex usage will stay the same. We’re also going to continue offering extended limits for dots for now, and usage of your dot will still continue to be separate from your default usage. And we'll of course also keep shipping better models and other efficiency improvements that increase your effective usage over time.
About five hours after that post, Tibor Blaho, whose X profile lists him as lead engineer at AIPRM, which makes a tool for ChatGPT and Claude users, posted a new disclaimer he had found in ChatGPT's web checkout. OpenAI's help article put the change in writing two days later.
Post on X by @btibor91
ChatGPT web app now has a new hidden shared usage limit checkout disclaimer - "Starting in December, plan usage will have a shared limit across all of ChatGPT and Codex"
Before the December change was confirmed, some developers counted the plans as good value. Theo, whose X profile lists him as chief executive of T3 Chat, reckoned after the October 29 changes that the $200 Pro plan "nets to ~$4.3k/month of usage," about 21 times its price on his figures.
Post on X by @theo
Ran the numbers on the plan changes for Codex, looking pretty good! $500 plan nets to ~$10.5k/month of usage $200 plan nets to ~$4.3k/month of usage
The objection is that coding and chatting now compete. Pankaj Kumar, a developer, posted at 3:49 p.m. ET on October 31, as the help article spread on X, that heavy Codex use now means less left for ChatGPT, "and vice versa."
Post on X by @pankajkumar_dev
No way OpenAI is actually doing this 💀 Starting in December, ChatGPT and Codex will share the SAME usage limits for Plus, Pro, and Business users. Both will consume a shared token allowance. Heavy Codex usage means less usage left for ChatGPT, and vice versa. Instant chat won't count toward the limit. So basically, no more separate limits for ChatGPT and Codex.
SaasBuzz found no post about the December change from OpenAI or its staff on X by 6:40 p.m. ET on October 31. The help article and the checkout line are the company's only public word on it so far.
What it means for developers and teams paying for Codex
For a Plus or Business user who codes in Codex all afternoon, the evening's ChatGPT work now spends from the same balance, unless it runs on Instant. On OpenAI's own ranges, model choice moves how far that balance goes by more than 60 times, so the model picker becomes the main cost control. OpenAI's guide already recommends its Luna model for "focused or repetitive tasks" and GPT-6 Astra for "especially demanding work."
Anthropic made the same choice earlier. Its Claude Code help page says its plan limits "are shared across Claude and Claude Code, meaning all activity in both tools counts against the same limits." Where the two now differ is the clock: Anthropic dropped the five-hour limit for verified Ultra accounts on October 30, while OpenAI has not said what happens to its windows.
The unit is the larger shift. Counting tokens instead of messages prices a long agent run and a one-line question by what each consumes, rather than as one message apiece. Enterprise customers on token-based agreements already pay by use: OpenAI's Codex article says usage on token-based Enterprise agreements "is billed in US dollars (USD), instead of deducted from credits." For a team lead budgeting Plus or Business seats, how far a seat goes now depends on the models and tasks people run, a number OpenAI has not yet published.
