OpenAI retools pricing mix

OpenAI quietly changed what happens when paying ChatGPT users hit the ceiling on its heavier reasoning model: instead of stopping cold, Plus and Pro accounts now get routed to GPT‑5.4 mini as a backup for GPT‑5.4 Thinking during high usage. (help.openai.com) That sounds small until you translate it into product terms: OpenAI is no longer treating “the best model” as a fixed entitlement for paid users, but as a premium lane that can spill over into a cheaper lane when demand spikes. (help.openai.com) OpenAI spelled this out in its March 18, 2026 model notes, saying GPT‑5.4 mini is not a model you manually pick from the menu for paid fallback use; it turns on when GPT‑5.4 Thinking rate limits are reached. (help.openai.com) At the same time, the company changed how it charges for Codex, its coding agent inside ChatGPT. On April 2, 2026, OpenAI said new and existing ChatGPT Business customers and new ChatGPT Enterprise customers would move to pricing tied to application programming interface token usage instead of per-message pricing. (help.openai.com) A token is the unit these systems bill on, like paying for electricity by the kilowatt-hour instead of paying one flat fee every time you flip a light switch. OpenAI’s Codex rate card says that shift applies first to Business and new Enterprise plans, while some existing Plus, Pro, Enterprise, and Education customers remain on legacy pricing until migration in the coming weeks. (help.openai.com) OpenAI also launched a new “Codex seat” for ChatGPT Business on April 2, 2026. That seat has no fixed monthly price, gives access to Codex only rather than the full ChatGPT workspace, and bills against workspace credits as people use it. (help.openai.com) In a separate OpenAI post, the company said Codex now has pay-as-you-go pricing for ChatGPT Business and Enterprise teams, and it framed that as a way for companies to “start and scale adoption” without committing to a bigger fixed contract first. (openai.com) That is the real pricing story here: OpenAI is slicing the product into two meters at once. One meter controls model quality under load inside ChatGPT, and the other meter charges coding help more like cloud infrastructure than like a bundled chat feature. (help.openai.com 1) (help.openai.com 2) The outside report that pushed this into sharper focus came from Windows Report on April 10, 2026, which said OpenAI is testing a $100-per-month ChatGPT Pro plan with 5 times more Codex usage and access to GPT‑5.4 Pro for heavy users. OpenAI has not published that specific $100 consumer plan on its own help pages in the sources reviewed here. (windowsreport.com) (help.openai.com) If that test becomes official, it would fit the same pattern as the fallback change and the Codex rate card: casual users get continuity, power users get tiers, and software teams get bills that look more like metered developer tools than like all-you-can-eat subscriptions. (windowsreport.com) (openai.com)