OpenAI has launched GPT-5.4, an advanced AI model designed to enhance knowledge work. The update features significant improvements in visual understanding, capable of analyzing images up to 10.24 million pixels or with a maximum dimension of 6,000 pixels. OpenAI reports that this iteration is 18 percent less prone to factual errors compared to previous versions.
This development comes amid competitive pressures, with rival Anthropic reportedly gaining users following a public dispute between the Trump administration and Anthropic regarding military applications of AI. Anthropic has since expanded access to its memory features for free users and reported a surge in new sign-ups.
OpenAI's strategy with GPT-5.4 focuses on maintaining its market position through enhanced capability and cost efficiency. The new model is now available across ChatGPT web and native apps, Codex, and the API. Subscribers to Plus, Team, and Pro tiers will receive GPT-5.4 Thinking, with GPT-5.4 Pro rolling out to the API, Edu, and Enterprise platforms.