OpenAI has launched GPT-5.4, the latest foundation model tailored for professional and technical tasks. This version enhances reasoning, coding skills, and workflow automation, presenting a unified system for real-world applications. GPT-5.4 comes in multiple variations, including a standard API model and the reasoning-focused GPT-5.4 Thinking in ChatGPT, along with a high-performance Pro version for complex workloads. The model excels in key benchmarks like OSWorld-Verified and APEX-Agents, showcasing its efficiency in document handling and multi-step workflows while reducing token usage. Notable features include Tool Search, which optimizes token consumption, and enhanced direct computer operation capabilities. OpenAI emphasizes improvements in accuracy, with GPT-5.4 being 33% less likely to produce errors. With safety evaluations indicating reduced chances of deception, GPT-5.4 aims to deliver reliable performance while cutting costs, making it an exceptional tool for developers and enterprises in diverse fields.
Source link
Share
Read more