OpenAI upgrades Codex & pricing

OpenAI has started charging some Codex users by consumption instead of fixed seats, while expanding the coding agent’s reach across desktop work. (openai.com) On April 2, OpenAI said ChatGPT Business and ChatGPT Enterprise customers can add “Codex-only” seats with no fixed seat fee, then pay based on token usage. The same post said ChatGPT Business’s annual price falls to $20 per seat from $25 for standard seats that still include capped Codex access. (openai.com) OpenAI’s help center says Business and Enterprise workspaces now have two seat types: standard ChatGPT seats and Codex-only seats. Business users keep per-seat limits on advanced features, but Codex-only seats and Enterprise usage draw from shared credits that admins can control. (help.openai.com) OpenAI also changed how Codex usage is measured on April 2. Its rate card says new and existing Business users, plus new Enterprise users, are now billed on input, cached input, and output tokens instead of per-message estimates. (help.openai.com) That pricing shift lands as OpenAI pushes Codex from a code generator into a software agent that can juggle longer jobs. OpenAI said in February that the Codex app lets multiple agents work in parallel on separate copies of the same repository, first on macOS and, as of March 4, on Windows too. (openai.com) By April, OpenAI said teams could start pilots without buying a full block of seats first. The company said more than 9 million paying business users rely on ChatGPT for work, more than 2 million builders use Codex each week, and Codex usage inside Business and Enterprise has grown sixfold since January. (openai.com) TechCrunch reported on April 16 that OpenAI had also expanded Codex’s ability to work on a user’s desktop in the background, opening apps and clicking through tasks while the person keeps using the machine. TechCrunch said OpenAI was positioning the update against Anthropic’s Claude Code, which has gained traction with business users. (techcrunch.com) OpenAI paired the Codex changes with a second product push on April 16: GPT-Rosalind, a reasoning model for biology, drug discovery, and translational medicine. OpenAI said the model is available as a research preview in ChatGPT, Codex, and the application programming interface for qualified customers through its trusted access program. (openai.com) OpenAI said GPT-Rosalind is built to help with literature review, hypothesis generation, experimental planning, and data analysis across chemistry, protein engineering, and genomics. The company said it is working with Amgen, Moderna, the Allen Institute, and Thermo Fisher Scientific on research workflows tied to discovery work that can take 10 to 15 years to reach regulatory approval in the United States. (openai.com) The combined message is that OpenAI is selling more than a chatbot seat. It is packaging coding agents and specialist models as metered tools that companies can test in small groups, then scale if the usage justifies the bill. (openai.com)