OpenAI releases the latest model GPT-5.4

The United States Open Artificial Intelligence Research Center announced on the 5th the release of the latest upgrade version of the artificial intelligence model GPT-5, called GPT-5.4, which is one of the most powerful and efficient cutting-edge models currently available for professional work. According to OpenAI, GPT-5.4 integrates the latest technological achievements in reasoning, programming, and intelligent agent workflows. The new model inherits the programming capabilities of the previous GPT-5.3-Codex and has been optimized for tasks such as handling tables, presentations, and documents, allowing the model to efficiently complete complex work. The model also significantly improves its ability to process long texts. The company stated that GPT-5.4 offers two models, Thinking and Pro. The Pro model is mainly aimed at users who need to achieve maximum performance in complex tasks. The Thinking model can provide a thinking plan before generating answers, allowing users to adjust the task direction during the generation process, reducing multi-round conversations and improving the matching of results with requirements. The Thinking model also enhances deep network research capabilities, providing higher quality, faster answers that better fit the task. OpenAI stated that GPT-5.4 is its first universal model that can directly operate computers, excelling not only in writing code but also issuing mouse and keyboard operation commands based on screen shots, allowing intelligent agents to operate computers and perform complex workflows between different applications. OpenAI stated that GPT-5.4 outperformed previous versions in multiple evaluations and has also improved in the accuracy of facts. The company announced that the old version of the GPT-5.2 Thinking model will be gradually phased out in the coming months.