OpenAI has introduced GPT 5.4, a new artificial intelligence model designed to support professional work across multiple digital environments. The model is now being rolled out in ChatGPT as GPT 5.4 Thinking and is also available through the company’s application programming interface and Codex. According to OpenAI, the release combines the latest improvements in reasoning, coding capabilities and automated workflows into a single frontier model aimed at developers, enterprises and professionals who rely on artificial intelligence tools to complete complex tasks.
GPT 5.4 builds on the company’s previous models by integrating the coding strengths of GPT 5.3 Codex with broader knowledge work capabilities. The system is designed to work more effectively across digital tools including spreadsheets, documents and presentations, allowing users to generate and edit structured content with greater accuracy. OpenAI said the model delivers improved performance in professional scenarios such as financial modelling, legal documentation and business analysis. Internal evaluations show GPT 5.4 achieving an 83 percent score on GDPval, a benchmark that measures the ability of artificial intelligence systems to produce real world work outputs across multiple industries, compared with about 71 percent recorded by the earlier GPT 5.2 model.
Another key development is the introduction of native computer use capabilities, allowing artificial intelligence agents to interact with software environments and websites. GPT 5.4 can perform tasks such as operating applications, browsing websites and executing workflows across systems through keyboard and mouse style interactions guided by screenshots. OpenAI stated that the model achieved a 75 percent success rate on the OSWorld Verified benchmark, which evaluates how effectively artificial intelligence systems can navigate computer interfaces. The model also supports up to one million tokens of context, enabling it to process and maintain large volumes of information while carrying out long and complex tasks.
The company also reported improvements in accuracy, search and tool integration. GPT 5.4 is designed to reduce factual errors while performing more persistent web searches and integrating with external tools more efficiently. In benchmarks measuring agent based browsing, the model demonstrated stronger performance in locating and synthesizing information from multiple online sources. OpenAI said the system is being deployed with expanded safety safeguards and monitoring frameworks as part of its preparedness policies for advanced artificial intelligence systems. GPT 5.4 is currently available to paid ChatGPT users including Plus, Team and Pro plans, while developers can access the model through OpenAI’s application programming interface with updated pricing that reflects its enhanced capabilities and efficiency.
Follow the SPIN IDG WhatsApp Channel for updates across the Smart Pakistan Insights Network covering all of Pakistan’s technology ecosystem.
