OpenAI reworks pricing

OpenAI just split one product stack into two billing logics at once: chat plans now have more steps, and Codex now charges like a meter instead of a bucket. On April 2, 2026, OpenAI switched Codex for Plus, Pro, Business, and new Enterprise customers from per-message estimates to token-based pricing. (help.openai.com) A token is a small chunk of text, so token pricing works like paying for electricity by kilowatt-hour instead of paying one flat fee per room. OpenAI’s new Codex card lists separate prices for input tokens, cached input tokens, and output tokens, with GPT-5.4 set at 62.50 credits per 1 million input tokens and 375 credits per 1 million output tokens. (help.openai.com) Cached input tokens are the reused parts of a job, like copying the same blueprint into every work order instead of redrawing it each time. OpenAI prices those reused tokens much lower, with GPT-5.4 cached input at 6.250 credits per 1 million tokens, or one-tenth of the regular input rate. (help.openai.com) The pricing table also tells developers which models are expensive and which are meant for cheaper background work. GPT-5.1-Codex-mini is listed at 6.25 credits per 1 million input tokens and 50 credits per 1 million output tokens, while GPT-5.3-Codex and GPT-5.2-Codex are both 43.75 credits for input and 350 credits for output. (help.openai.com) OpenAI also says Fast mode burns credits twice as quickly, which turns speed into a direct budget choice. The company’s own help page says average Codex spend lands around $100 to $200 per developer per month, with big swings depending on model choice, parallel agents, automations, and Fast mode. (help.openai.com) At the same time, OpenAI changed the consumer ladder above ChatGPT Plus. Its Pro help page now shows two Pro options: a $100 plan with 5 times the usage of Plus and a $200 plan with 20 times the usage of Plus. (help.openai.com) That means the old top-tier Pro plan did not disappear; it moved up and became the heavier-duty version. OpenAI says both Pro plans include the same core capabilities, but the $100 tier is for “real projects” while the $200 tier is for running demanding workflows continuously across parallel projects. (help.openai.com) Codex sits in the middle of both changes because it is no longer just a feature hidden inside ChatGPT chat windows. OpenAI describes Codex as a coding agent that can edit files, run commands, execute tests, work in Visual Studio Code, Cursor, and Windsurf, and run multiple agents in parallel in a separate app. (help.openai.com) Business customers got a parallel reshuffle. OpenAI says ChatGPT Business now has two seat types as of April 2, 2026: a standard ChatGPT seat and a Codex-only seat, and the standard Business seat price was cut by $5 to $25 monthly or $20 on annual billing. (help.openai.com 1) (help.openai.com 2) The new Codex-only seat has no fixed monthly charge, which makes it look less like a software license and more like a cloud compute tab. OpenAI says those seats are usage-based, require workspace credits for activity, and can be added with no minimum seat count, while standard ChatGPT seats still require at least two users. (help.openai.com) Put together, the pricing rewrite draws a cleaner line between people who mainly chat and people who mainly ship code. OpenAI is keeping ChatGPT as a subscription product, but it is pushing Codex toward the economics of an infrastructure service where every prompt, cache hit, and generated line has a visible meter attached. (help.openai.com 1) (help.openai.com 2)