GPT-5.4 to bring a million-token context window and an extreme reasoning mode (the-decoder.com)

OpenAI is set to announce GPT-5.4, which reportedly features a groundbreaking one-million-token context window, significantly surpassing the 400,000 tokens available in the current GPT-5.2 version. This enhancement positions OpenAI competitively alongside industry rivals like Google and Anthropic, offering improved performance in long-duration tasks, particularly beneficial for applications such as Codex, OpenAI's programming tool. The model is anticipated to reduce error rates, thereby increasing reliability for complex, extended interactions. Additionally, GPT-5.4 will introduce an "extreme reasoning mode," enabling the model to allocate more computational resources for difficult queries, which targets research professionals rather than casual users seeking quick responses. This feature underscores OpenAI's strategy to balance user expectations and mitigate the hype often surrounding new releases, following a period of slower-than-expected user growth. With these advancements, GPT-5.4 aims to enhance the capabilities and versatility of AI applications, marking a significant step forward in the evolving landscape of AI/ML technology.