OpenAI formalises Codex
OpenAI has turned its coding assistant into a priced product and reset usage limits after the tool hit roughly 3 million weekly users, signalling it’s past the experimental stage and into commercialisation ( ). The company published a Codex rate card and plans a model lineup rollout (GPT‑5.2-codex, GPT‑5.1-codex mini/max and related variants) next week, which treats coding models as a tiered SaaS offering rather than a single research artifact (businesstoday.in).
OpenAI’s coding tool just crossed 3 million weekly users, and the company responded the way software companies do when a product stops being a demo and starts being infrastructure: it reset usage limits and published a formal rate card. (openai.com) (businesstoday.in) (help.openai.com) Codex is OpenAI’s coding agent, which means it does more than suggest one line at a time. OpenAI says it can navigate a code repository, edit files, run commands, execute tests, review pull requests, and work in the cloud while a developer keeps doing something else. (help.openai.com) That shift matters because old coding assistants behaved like autocomplete, while this one is sold more like a contractor with a meter running in the background. OpenAI now describes Codex in terms of clients, sandboxes, automations, GitHub reviews, and parallel agents across projects. (help.openai.com) On April 2, 2026, OpenAI changed Codex pricing for ChatGPT Business and new Enterprise customers from per-message billing to token-based billing. Tokens are the small chunks of text a model reads and writes, so the new system charges separately for input, cached input, and output instead of treating every request like the same-sized job. (help.openai.com) The published menu now looks like a cloud software price sheet. OpenAI lists models including GPT-5.4, GPT-5.4 Mini, GPT-5.3 Codex, GPT-5.2 Codex, GPT-5.1 Codex Max, and GPT-5.1 Codex mini, with different credit costs for each million tokens. (help.openai.com) The cheapest listed coding model on that card is GPT-5.1 Codex mini at 6.25 credits per million input tokens, while GPT-5.4 is listed at 62.50 credits for the same input volume and 375 credits per million output tokens. OpenAI also says Fast mode burns credits at double the normal rate. (help.openai.com) OpenAI is also telling customers to think about Codex as a recurring budget line, not a novelty feature. Its help page says average Codex usage runs about $100 to $200 per developer per month, with big swings depending on model choice, automations, the number of running instances, and Fast mode. (help.openai.com) The usage reset was tied directly to growth. Business Today reported that Sam Altman said OpenAI would reset Codex limits at 3 million weekly users and repeat that at every additional million users until the product reaches 10 million. (businesstoday.in) That growth has been fast even by OpenAI standards. Business Today reported that Codex head Thibault Sottiaux said weekly users rose from 2 million to 3 million in a little under a month. (businesstoday.in) The product is also spreading across more places where developers already work. OpenAI says Codex can be used through a command line interface, integrated development environment extensions for Visual Studio Code, Cursor, and Windsurf, a desktop app for MacOS and Windows, GitHub code review, and even Slack. (help.openai.com) OpenAI’s own enterprise note on April 8 put Codex’s 3 million weekly active users next to another number: its application programming interfaces now process more than 15 billion tokens per minute. That is the language of a company telling customers that coding agents are no longer a side experiment inside chat, but one more metered service in a broader enterprise stack. (openai.com)