OpenAI rolls out GPT‑5.4

OpenAI did not just drop a new chatbot on March 5, 2026. It released GPT‑5.4 into ChatGPT, Codex, and the application programming interface on the same day, which means the launch was aimed at three products people already plug into work, not one demo people try for fun. (openai.com) OpenAI says GPT‑5.4 is built for “professional work,” and the company tied that claim to coding, computer use, tool search, and a context window of up to 1 million tokens in Codex and the application programming interface. A context window is the model’s working memory, and 1 million tokens is roughly enough room for very long codebases, document sets, or multi-step agent plans. (openai.com, developers.openai.com) The new piece here is computer use. OpenAI says GPT‑5.4 is its first general-purpose model with native computer-use capabilities in Codex and the application programming interface, which means the model can operate software more like a remote worker clicking through screens than a chatbot waiting for one prompt at a time. (openai.com) ChatGPT got a different version of the same push. OpenAI says GPT‑5.4 appears there as GPT‑5.4 Thinking, and it can show an upfront plan before finishing a long answer so a user can steer the job midstream instead of rewriting the request after the fact. (help.openai.com, openai.com) This came after OpenAI cleaned out older chat models on February 13, 2026. The company says GPT‑4o, GPT‑4.1, GPT‑4.1 mini, OpenAI o4‑mini, and GPT‑5 Instant and Thinking were retired from ChatGPT, which makes GPT‑5.3 and GPT‑5.4 the center of the current lineup rather than one more option in a crowded menu. (help.openai.com, help.openai.com) Codex is where the business shift becomes easiest to see. OpenAI updated its Codex pricing on April 2, 2026 to align with application programming interface token usage instead of per-message pricing for new ChatGPT Business plans and new ChatGPT Enterprise plans, which turns coding help into a metered software bill instead of an all-you-can-eat chat perk. (help.openai.com, help.openai.com) OpenAI also added a new Codex-only seat for ChatGPT Business and ChatGPT Enterprise on April 2, 2026. That seat has no fixed monthly user fee and charges by usage, which lets a company give a developer access to coding agents without buying that person the full ChatGPT workspace. (help.openai.com, help.openai.com, help.openai.com) The application programming interface side tells the same story in numbers. OpenAI lists GPT‑5.4 at $2.50 per 1 million input tokens, $0.25 per 1 million cached input tokens, and $15.00 per 1 million output tokens in Batch application programming interface pricing, with extra charges for very large prompts above 272,000 input tokens. (developers.openai.com) Smaller versions followed fast. OpenAI introduced GPT‑5.4 mini and GPT‑5.4 nano later in March 2026 and said they were optimized for coding, tool use, multimodal reasoning, and high-volume application programming interface and sub-agent workloads, which is the language of teams building systems of many models rather than one assistant in one chat box. (openai.com) OpenAI’s release notes also show Custom Actions appearing for paid users in ChatGPT, alongside connectors and app updates for tools like Box, Notion, Linear, and Dropbox. Put together with GPT‑5.4, the picture is less “ask me anything” and more “let me work inside the software your company already uses.” (help.openai.com, help.openai.com) That is why this rollout looks different from the model launches of 2023 and 2024. OpenAI is now selling a stack with model tiers, usage meters, seat types, connectors, and agent features, and GPT‑5.4 is the model release that makes that platform strategy visible all at once. (openai.com, help.openai.com, developers.openai.com)