OpenAI Launches GPT-5.4, Expands Capabilities
OpenAI launched GPT-5.4, boasting enhanced coding, writing, and reasoning. The model offers expanded context windows and interruptible reasoning, allowing users to pause or redirect the AI mid-response, positioning it as a major upgrade for professional knowledge work. However, early testers noted some reliability gaps, emphasizing the need for human oversight.
GPT-5.4 boasts a 1.05 million token context window, enabling it to handle extensive datasets like large code repositories or years of historical logs. This expanded context allows for more complex, multi-step reasoning while maintaining accuracy. The model combines the reasoning improvements of GPT-5.2 with the coding strengths of GPT-5.3 Codex. GPT-5.4 introduces native computer use capabilities, allowing it to directly control a user's browser and desktop. In testing, GPT-5.4 achieved a 75% score on the OSWorld benchmark, surpassing the human baseline of 72.4%. OpenAI says that, compared to GPT-5.2, GPT-5.4 is 33% less likely to make false claims and 18% less likely to contain factual errors. GPT-5.4 is rolling out across ChatGPT, Codex, and the API, with a higher-performance "Pro" version available for more complex tasks. Within ChatGPT, "GPT-5.4 Thinking" replaces "GPT-5.2 Thinking" for Plus, Team, and Pro users. OpenAI is pricing GPT-5.4 higher per token than GPT-5.2, but claims increased token efficiency will offset the cost. The release of GPT-5.4 coincides with debates surrounding OpenAI's partnerships, including a contract with the U.S. Department of Defense. OpenAI is also investing heavily in infrastructure, including data centers, with plans for facilities in Texas, New Mexico, Ohio, Michigan and Wisconsin. Sam Altman has described GPT-5.4 as his "favorite model to talk to".
