OpenAI Launches GPT-5.4 Agent
OpenAI rolled out GPT-5.4 with "agentic powers" designed to work directly on computers as a personal assistant. The model can control computers, automate repetitive digital tasks, and sift through large volumes of information. Early reviews highlight its versatility for automating everything from email sorting to document research, potentially reshaping daily work and personal productivity.
The move toward agentic AI marks a significant industry shift from conversational partners to autonomous systems that execute tasks. This evolution addresses the growing demand for AI that doesn't just answer questions but actively completes multi-step objectives with minimal human oversight. OpenAI's GPT-5.4 streamlines its product line by integrating reasoning and coding capabilities that were previously separated in models like GPT-5.2 Thinking and GPT-5.3 Codex. This consolidation simplifies the user experience, offering a single, more powerful model for a wider range of tasks. The core of its new power is a native "Computer Use" feature, allowing the AI to interpret a computer's screen and directly control the mouse and keyboard. This enables it to navigate between different applications to complete complex workflows, a key step towards a true AI agent. On the OSWorld-Verified benchmark, which tests an AI's ability to navigate desktop environments, GPT-5.4 scored 75%, significantly outperforming both its predecessor GPT-5.2 (47.3%) and the average human result (72.4%). The model's context window has been massively expanded, handling up to 1 million tokens, which allows it to maintain context during long, complex projects. OpenAI also claims GPT-5.4 is its most factual model yet, with individual claims being 33% less likely to be false compared to GPT-5.2. This launch intensifies the race for agentic AI dominance, placing OpenAI in direct competition with rivals like Google and Anthropic, who are also heavily invested in developing autonomous AI systems. The industry is rapidly moving beyond chatbots to create "digital workers" that can be integrated into everyday business operations.
