OpenAI spreads GPT‑5.4 across products
OpenAI is rolling GPT‑5.4 into ChatGPT, Codex and its API, embedding the model across chat, developer tools and paid tiers while adjusting Codex pricing and credits for different plans. That rollout shows model releases are now strategic commercial levers across consumer, pro and enterprise surfaces, and OpenAI has also signalled retail IPO allocation plans for investors even as demand remains strong ( ).
OpenAI did not ship GPT‑5.4 as a single chatbot update. On March 5, 2026, it pushed the model into ChatGPT, the application programming interface for developers, and Codex, its coding product, at the same time. (openai.com) In ChatGPT, the release arrived as “GPT‑5.4 Thinking,” and OpenAI also added “GPT‑5.4 Pro” for users who want a slower, higher-end version for harder tasks. The company’s pricing page now says paid plans include GPT‑5.4 access, with Pro getting access to GPT‑5.4 Pro as well. (openai.com, chatgpt.com) In the application programming interface, GPT‑5.4 is priced at $2.50 per 1 million input tokens and $15 per 1 million output tokens. GPT‑5.4 Pro is much pricier at $30 input and $180 output, which turns the model ladder into a menu for different budgets as much as different tasks. (developers.openai.com, developers.openai.com) OpenAI is also selling the same family through Codex, which is the company’s tool for software work across the web app, command line interface, integrated development environment extension, and iPhone app. The Codex pricing page says Plus includes the latest models including GPT‑5.4 and GPT‑5.3‑Codex, while higher tiers raise limits and let teams add credits. (developers.openai.com) That is a change from the old pattern where a coding model could live in its own lane. OpenAI’s current model guide says GPT‑5.4 is now the default starting point for both general work and most coding tasks, which folds chat, coding, and workflow automation into one flagship product. (developers.openai.com) The company is still keeping a specialist option on the shelf. OpenAI’s Codex models page says GPT‑5.3‑Codex‑Spark remains available in research preview for ChatGPT Pro subscribers and is tuned for near-instant coding iteration. (developers.openai.com, help.openai.com) The technical pitch for GPT‑5.4 is not just that it is smarter. OpenAI says the model has a 1.05 million token context window in the application programming interface and Codex, which means it can keep far more code, documents, or instructions in view at once than older systems. (developers.openai.com, openai.com) OpenAI is also arguing that a higher sticker price can still mean a lower bill. Its GPT‑5.4 guide says the model produces higher-quality work with fewer iterations, and the launch post calls it the company’s most capable and efficient frontier model for professional work. (developers.openai.com, openai.com) So the release is really three launches tied together: a consumer launch inside ChatGPT, a developer launch in the application programming interface, and a workplace launch in Codex. One model name now moves subscription value, usage revenue, and enterprise seat pricing at the same time. (openai.com, chatgpt.com, developers.openai.com) OpenAI has paired that product bundling with a capital-markets signal. In a video posted on X, chief executive Sam Altman said the company wants to reserve some future initial public offering allocation for retail users, even though he also said demand for OpenAI stock is already intense. (x.com) Put together, GPT‑5.4 is not being treated like a lab demo. It is being used as the same underlying engine for chat, code, and paid developer infrastructure, with different limits, prices, and plan perks wrapped around it. (openai.com, developers.openai.com, developers.openai.com)
