OpenAI ties Codex to agent credits

OpenAI has updated its Help Center pricing pages to tie Codex more directly to the company’s broader agent products. The Codex rate card now says usage from Codex, ChatGPT Work, ChatGPT for Excel and Workspace Agents can draw from the same agentic usage and credit pool when those features are included in a plan. That change puts coding work inside the same metered bucket as other agent-driven tasks in ChatGPT. A separate OpenAI help article describes ChatGPT Work as an agent for longer, multi-step work, while Codex is positioned for software development and technical tasks. ### Which OpenAI products are now counted together? OpenAI’s Codex rate card says “usage from Codex, ChatGPT Work, ChatGPT for Excel, and Workspace Agents draws from the same agentic usage and credit pool” when those features are available on a user’s plan. (help.openai.com) The page says task consumption varies with job size, model choice and where the task runs. OpenAI’s enterprise token-pricing page uses a similar grouping, saying usage across Codex, ChatGPT Work, ChatGPT for Excel, ChatGPT for PowerPoint and Workspace Agents is charged using the applicable model and feature rates in that article. (help.openai.com) That indicates OpenAI is aligning coding and office-agent features under a shared usage framework, though the billing unit differs by contract type. (help.openai.com) ### What changed in the Codex pricing page itself? OpenAI’s Codex rate card says it outlines “current credit rates” for Plus, Pro, Business, Enterprise, Edu, Health and Gov plans under a flexible pricing structure. The article was updated two days before it was crawled by the search tool. The same page says OpenAI moved Codex pricing to API token-based billing on April 2, 2026 for Plus, Pro and Business users and for new ChatGPT Enterprise customers, then extended that update on April 23, 2026 to existing Enterprise plans including Edu, Health, Gov and ChatGPT for Teachers. (help.openai.com) That replaced older per-message estimates with token-based charging for input, cached input and output. (help.openai.com) ### What is OpenAI saying about GPT-5.3-Codex-Spark? OpenAI lists GPT-5.3-Codex-Spark on the Codex rate card as a “research preview,” and the page says its credit rates are not final. The same article says Codex code review uses GPT-5.3-Codex by default. That matters because buyers do not yet have a settled number for that model in the published rate card. OpenAI also says users should check separate rate-card pages for plan-specific pricing and effective dates, depending on whether they are on Business, Enterprise or consumer plans. (help.openai.com) ### How does this look for enterprise customers? OpenAI’s enterprise token-pricing page says it applies only to new Enterprise customers whose agreements specify usage-based billing in U.S. dollars. (help.openai.com) The page says rates are charged per 1 million input, cached input and output tokens unless another billing unit is listed. Existing Business and many Enterprise or Edu customers are pointed instead to the credit-based ChatGPT and Codex rate cards. (help.openai.com) OpenAI says customers who are unsure which rate card applies should contact their OpenAI representative. ### Where can users see what they are spending? OpenAI’s Codex documentation says users can monitor limits in the Codex settings usage panel. (help.openai.com) Depending on plan and workspace role, the page says users may also be able to see remaining credits, buy credits or manage auto-recharge settings there. OpenAI’s current documentation leaves the next concrete checkpoints on its own pricing pages. (help.openai.com) The Codex rate card says GPT-5.4 and GPT-5.4 mini in Codex are scheduled to retire for ChatGPT-login users on August 31, 2026, with GPT-5.6 Terra and GPT-5.6 Luna listed as replacements. (help.openai.com)