ChatGPT plans move to one token budget for chat and Codex in December, and Claude's $200 plan is now Ultra. Budget AI work in tokens

Both big AI subscriptions changed how they count your usage in the past few days, and the direction is clear: you are buying a budget, not unlimited prompts.

What happened

OpenAI is folding its Chat and Work modes into one ChatGPT interface. According to an updated OpenAI support page reported by The Decoder, from December the Plus, Pro and Business plans will share a single token budget across ChatGPT and Codex. More complex models and tasks use the budget faster. Chats with Instant models do not count. You can check your remaining balance and reset times in Settings, and when you hit the limit you can wait, upgrade or buy extra credits.

OpenAI also added card and device limits on Go, Plus and the Pro plans. Logging out of an account does not free up a spot for another one. Business, Enterprise and Edu plans are not counted.

On Friday, Anthropic renamed its plans. Max 5x is now Max at $100 a month and Max 20x is now Ultra at $200. Verified Ultra accounts no longer have the limit that resets every five hours, but weekly limits stay. The Next Web notes the pricing page now lists Ultra at ten times Pro usage a week, and Ultra includes $200 a month in API credits.

My take

For anyone building automations for clients, this settles an old shortcut. A personal chat subscription was never the right engine for a client workflow, and now it is a metered pool shared with your own daily work. Run a heavy Codex session in the morning and your afternoon research gets throttled.

What I do instead:

1. Client automations run on API keys, one per client or project, so the cost lands on the right invoice.
2. Routine steps like classification or reformatting go to the cheapest model that passes the test. The expensive model only handles the step that needs judgment.
3. Seat plans are for people. Track who on the team burns through the budget and why before you buy more seats.

Treat tokens like any other operating cost. Measure them per workflow, and the pricing changes stop being a surprise.