OpenAI updates Codex rate card
OpenAI published an updated Codex rate card that structures code-model usage by plans and credit tiers across Plus, Pro, Business and Enterprise/Edu offerings. The Help Center posting lays out explicit pricing layers for different user tiers and usage patterns. (help.openai.com)
OpenAI has rewritten how Codex is billed, shifting many ChatGPT users from per-message pricing to token-based pricing as of April 2, 2026. (help.openai.com) The new rate card applies to new and existing ChatGPT Plus and Pro customers, new and existing ChatGPT Business customers, and new ChatGPT Enterprise customers, according to an OpenAI Help Center article updated about 19 hours ago. Existing Enterprise customers, along with ChatGPT Edu, Teacher, and Healthcare plans, remain on a legacy rate card for now. (help.openai.com) Under the new system, Codex usage is priced by credits per 1 million tokens, split into input, cached input, and output tokens. OpenAI lists GPT-5.4 at 62.50 credits for input tokens and 375 credits for output tokens, while GPT-5.1-Codex-mini is priced at 6.25 credits for input and 50 credits for output; Fast mode uses 2 times as many credits. (help.openai.com) OpenAI says the change replaces “average per-message estimates” with a direct mapping to token usage, so the bill now depends on how much code and context a task actually consumes. The company says that makes pricing easier to track across long prompts, cached context, and large outputs. (help.openai.com) The update also sharpens the product ladder around Codex. On OpenAI’s Codex pricing page, Plus is listed at $20 a month, while Pro now starts at $100 a month with “10x or 20x higher rate limits than Plus,” and OpenAI says the current boost runs through May 31, 2026. (developers.openai.com) That page says the $100 Pro plan is currently 10 times Plus usage and the $200 Pro plan is currently 20 times Plus usage. It also says Plus includes Codex on the web, in the command line interface, in the integrated development environment extension, and on iOS, with access to models including GPT-5.4 and GPT-5.3-Codex. (developers.openai.com) For business customers, OpenAI has also added two seat types as of April 2, 2026: a standard ChatGPT seat and a Codex-only seat. The flexible pricing rules say Codex seats are available on ChatGPT Business and ChatGPT Enterprise, but not on Edu, Teachers, or Healthcare plans. (help.openai.com) Business users still get per-seat limits on advanced features, and they can keep using Codex after hitting those limits only if the workspace has purchased credits. Enterprise and Edu customers instead draw from a shared credit pool, with owners able to set spend controls by group. (help.openai.com) OpenAI’s Help Center says Codex now costs about $100 to $200 per developer per month on average, though it says actual usage varies widely by model, number of running instances, automations, and Fast mode. The company also says GPT-5.3-Codex-Spark may appear as a research preview and that its rates are not final. (help.openai.com) The practical effect is that OpenAI has moved Codex closer to the way its application programming interface has long charged for models: by the amount of text and code processed, not by the number of chats. For developers and workplace admins, the new rate card puts a number on how much context, output, and speed settings can change the bill. (help.openai.com)