OpenAI posts Codex price card

OpenAI has put a clearer public price on Codex usage, posting a rate card that ties credits directly to token consumption across ChatGPT Plus, Pro, Business, Enterprise, Education, Health and Government plans. The Help Center page says the company changed Codex pricing on April 2, 2026, moving away from per-message estimates for new and existing Plus, Pro and Business customers, and for new Enterprise plans. A second update on April 23 extended that model to existing Enterprise plans, including Edu, Health, Gov and ChatGPT for Teachers. The page was updated five days before May 14, according to OpenAI’s Help Center. The new card gives buyers a more granular way to estimate usage. OpenAI says Codex is now priced in credits per 1 million input tokens, cached input tokens and output tokens, replacing average message-based estimates with what it called a “direct mapping” between token usage and credits. For GPT-5.5, the published rate is 125 credits for input tokens, 12.50 for cached input tokens and 750 for output tokens per 1 million tokens. GPT-5.4 is listed at 62.50, 6.250 and 375 credits, while GPT-5.3-Codex and GPT-5.2 are each listed at 43.75, 4.375 and 350 credits. (help.openai.com) ### Which plans are covered by the new Codex pricing? OpenAI’s Help Center says the token-based rate card applies to new and existing ChatGPT Plus and Pro customers, new and existing ChatGPT Business customers, and new and existing Enterprise, Edu, Gov, Health and ChatGPT for Teachers customers. The company added that a “small subset” of Enterprise customers remains on a legacy rate card and should contact sales. (help.openai.com) The developers pricing page says Codex is included in ChatGPT Free, Go, Plus, Pro, Business, Edu and Enterprise plans, though feature access and limits vary by tier. Plus is listed at $20 a month. Pro starts at $100 a month, with OpenAI advertising “5x, 10x or 20x more Codex usage than Plus” and a temporary offer to double normal Codex usage on the $100 tier through May 31, 2026. Enterprise and Edu are sold through sales. (help.openai.com) ### What does the rate card say about actual monthly spend? OpenAI’s Help Center says Codex costs “~$100-$200/developer per month” on average, while adding that usage varies widely by model, the number of instances running, automations and use of fast mode. The same page says fast mode consumes credits at a higher rate for supported models, and that code review uses GPT-5.3-Codex. (developers.openai.com) The company also says customers can monitor remaining credit, buy more credit and manage auto-reload settings from the Codex usage panel. That matters because the published rates separate input, cached input and output, which means the same task can have different costs depending on how much context is reused and how much code the model generates. That last point is an inference from the pricing structure described by OpenAI. (help.openai.com) ### How does this line up with OpenAI’s pitch for GPT-5.5 Instant? OpenAI said on May 5 that GPT-5.5 Instant had become ChatGPT’s default model and described it as “smarter and more accurate,” with clearer and more concise answers and stronger personalization. The company said in internal evaluations that GPT-5.5 Instant produced 52.5% fewer hallucinated claims than GPT-5.3 Instant on high-stakes prompts, and 37.3% fewer inaccurate claims on challenging conversations users had flagged for factual errors. (help.openai.com) OpenAI’s developers pricing page also says GPT-5.5 “uses significantly fewer tokens” than GPT-5.4 to achieve comparable results and says those efficiency gains support higher usage limits. ### Did outside testing match those claims? The New Stack reported on May 13 that a head-to-head test of GPT-5.5 Instant against GPT-5.2 found mixed results. Jessica Wachtel wrote that OpenAI had made three specific claims around smarter and more accurate answers, more concise responses and deeper personalization. (openai.com) She reported that GPT-5.2 was “consistently more concise” in her tests, while GPT-5.5 was more thorough and conversational. (developers.openai.com) That article compared GPT-5.5 Instant with GPT-5.2 rather than GPT-5.3 Instant, which was OpenAI’s baseline in its own announcement. Even so, the comparison adds an external check on how the newer model behaves in ordinary use, separate from OpenAI’s internal evaluations. ### Where can teams check the numbers themselves? OpenAI has posted the Codex rate card in its Help Center and plan-level Codex pricing on its developer site. (thenewstack.io) The GPT-5.5 Instant product note remains on OpenAI’s site, and The New Stack’s May 13 test lays out the prompts and comparisons it used. OpenAI’s Pro promotion on the developer pricing page runs through May 31, 2026. (help.openai.com)