On March 5th (Thursday), local time, OpenAI introduced its latest foundational model, GPT-5.4, heralding it as the "most robust, efficient, and advanced model ever developed for professional tasks." Alongside the standard edition, OpenAI also launched two specialized variants: GPT-5.4 Thinking, which prioritizes intricate reasoning abilities, and GPT-5.4 Pro, engineered for high-performance environments. GPT-5.4 Thinking introduces a groundbreaking feature—'transparent display of the reasoning process'—empowering users to steer task directions in real-time as the model generates responses. This innovation also enhances contextual awareness during in-depth web searches and within extended logical sequences. GPT-5.4 Pro caters to enterprise-scale demands, boasting a context window capable of processing up to 1 million tokens. Compared to its predecessor, GPT-5.2, it achieves remarkable gains in reasoning efficiency, utilizing fewer tokens and delivering swifter responses when tackling complex challenges. Moreover, GPT-5.4 stands out as the first model to integrate native-level 'computer operation' capabilities, enabling agents to seamlessly navigate and execute intricate workflows across diverse software platforms. In computer operation proficiency tests, it scored 75%, outperforming the average human benchmark and surpassing several rival products. Presently, GPT-5.4 Thinking is accessible to ChatGPT Plus, Team, and Pro subscribers, whereas GPT-5.4 Pro is primarily targeted at enterprises and developers.