OpenAI has launched GPT-5.4, a new version of its artificial intelligence model, in ChatGPT. The company says the model is its “most capable and efficient frontier model for professional work”. The new AI model is available in ChatGPT as GPT-5.4 Thinking, and it is also being released through its developer API and the coding platform Codex. The update combines improvements in reasoning, coding and task automation into a single system. OpenAI said the new model is designed to help users complete tasks involving documents, spreadsheets, presentations and software tools. A second version called GPT-5.4 Pro is also being released for users who need higher performance on complex tasks.
OpenAI’s GPT-5.4: Key features
OpenAI says GPT-5.4 brings together improvements in reasoning, coding and workflow automation. The model builds on earlier systems and includes coding capabilities from GPT-5.3-Codex.In ChatGPT, GPT-5.4 Thinking can show an outline of its plan before generating a final answer. This allows users to change instructions while the model is working. The feature is designed to help users guide the response without restarting the conversation.The model also improves how ChatGPT searches the internet for information. According to OpenAI, it can handle longer questions and maintain context across multiple steps while collecting information from different sources.GPT-5.4 also introduces built-in computer-use abilities. This allows AI agents to interact with software, websites and tools to complete tasks such as managing files or working across applications. The system supports up to 1 million tokens of context, allowing it to process large documents and long conversations.OpenAI said the model also improves tasks involving professional work such as creating spreadsheets, writing documents and preparing presentations. The company reported that GPT-5.4 achieved 83% wins or ties against industry professionals in a benchmark called GDPval, which tests tasks across 44 occupations.The company also reported improvements in coding and tool usage. On the SWE-Bench Pro coding benchmark, GPT-5.4 scored 57.7%, slightly higher than earlier models.
OpenAI’s GPT-5.4: Availability
OpenAI said GPT-5.4 is rolling out gradually starting today.In ChatGPT, GPT-5.4 Thinking is available to Plus, Team and Pro users, replacing GPT-5.2 Thinking. The previous model will remain available for paid users for three months under a “Legacy Models” section before it is removed.Enterprise and education customers can enable early access through administrator settings. GPT-5.4 Pro is available to Pro and Enterprise plans.Developers can also access GPT-5.4 through the OpenAI API using the model name gpt-5.4, while gpt-5.4-pro is available for higher performance tasks.The model is also integrated into Codex, where developers can use it for coding and software development workflows. OpenAI said Codex users can also test an experimental feature that allows the system to interact with applications during development tasks.
