OpenAI has released GPT-5.4, its most powerful artificial intelligence (AI) model to date for professional use, combining advanced reasoning, coding, and the ability to autonomously operate computers and software applications within a single system.
The model, launched on Thursday, March 5, is available across ChatGPT, the OpenAI application programming interface (API), and Codex. GPT-5.4 Thinking replaces GPT-5.2 Thinking for ChatGPT Plus, Team, and Pro subscribers, with GPT-5.2 Thinking remaining accessible under Legacy Models until its retirement on June 5, 2026. A higher-performance tier, GPT-5.4 Pro, is available for Pro, Enterprise and Edu plan subscribers.
The new model unifies the coding strengths of GPT-5.3-Codex with improved reasoning and professional workflow capabilities, including record benchmark scores in computer use evaluations OSWorld-Verified and WebArena Verified, and a record 83 percent score on OpenAI’s GDPval knowledge work benchmark, measuring the model’s ability to match or exceed industry professionals.
GPT-5.4 is the first general-purpose model OpenAI has released with native computer-use capabilities, enabling it to autonomously navigate desktops, browsers, and software applications on behalf of users without requiring additional setup.
Through the API, the model supports context windows of up to one million tokens, the largest ever offered by OpenAI, allowing agents to plan, execute, and verify tasks across extended workflows. OpenAI also introduced Tool Search with this release, helping agents identify and use the right tools from large ecosystems of connectors more efficiently.
On accuracy, OpenAI said GPT-5.4’s individual claims are 33 percent less likely to be false and its full responses are 18 percent less likely to contain errors compared with GPT-5.2, making it the company’s most factual model to date.
In tests simulating junior investment banking analyst tasks, GPT-5.4 achieved an average score of 87.3 percent in spreadsheet modelling, compared with 68.4 percent for GPT-5.2, reflecting meaningful gains in professional finance applications.
The launch intensifies competition with rival Anthropic, which has made enterprise agentic tools a strategic focus since mid-2025. OpenAI is simultaneously navigating internal and public pressure following its decision to offer services to the United States Department of Defense, a move that reportedly cost the platform approximately 1.5 million users.
OpenAI also launched a ChatGPT for Excel add-in and updated spreadsheet and presentation productivity tools alongside the GPT-5.4 release.
