OpenAI launched GPT-5.4, a model explicitly designed for professional productivity tasks like coding and data analysis. It introduces native computer-use capabilities, allowing it to navigate desktop environments and operate multiple apps simultaneously via improved mouse and keyboard commands. Within ChatGPT, it becomes the default for Thinking mode, outlining its approach before generating a response and allowing user course-correction.

The model touts an 18% reduction in factual errors compared to GPT-5.2 and enhanced web research for specific queries. For API customers, OpenAI positions it as the most token-efficient reasoning model, but pricing reflects this power: one million input tokens costs $2.50, up from $1.75. This release is limited to Pro, Enterprise, and developer tiers, not free or Plus users.

This professional pivot follows competitive pressure; Microsoft added Anthropic's Claude to Copilot 365 after reports it excelled at spreadsheet and presentation tasks. With $25 billion in annualized revenue but massive data center costs and no profitability, OpenAI is betting on high-value productivity tools to build a sustainable business, moving beyond consumer chatbot features.