[Digital Today reporter Chi-gyu Hwang (황치규)] OpenAI will soon unveil the GPT-5.4 model, The Information reported on March 4, citing a source familiar with the matter.
GPT-5.4 is expected to offer a context window twice as large as the current GPT-5.2 model. It is expected to increase to 1 million tokens from about 400,000 tokens. That would be on par with Google and Anthropic. Google and Anthropic have already supported 1 million-token context windows in large language models. OpenAI supported a 1 million-token context in GPT-4.1, but not in 5.2.
A context window refers to the amount of input and output text that large language model systems can process while running. A larger context window means users can have a longer conversation with an AI chatbot at one time.
According to The Information, GPT-5.4 has improved performance on tasks that can take as long as several hours. That means it is better able to remember details requested by users and what it can and cannot do across multi-step processes. The Information said this could be useful for applications such as OpenAI Codex, a coding AI tool.
Chi-gyu Hwang
delight@d-today.co.kr