GPT-5.4 Leak Points to 1M Token Context Window

The Information reports that OpenAI’s next model iteration, GPT-5.4, will double its context window to one million tokens and add an “extreme” reasoning mode designed to throw significantly more compute at hard problems. The target audience is researchers, not everyday ChatGPT users. Context window size has quietly become the spec war nobody talks about – Google’s Gemini hit one million tokens over a year ago, and Anthropic has been pushing similar boundaries with Claude. OpenAI framing this as a research-grade feature rather than a consumer upgrade fits the pattern of their post-GPT-5 strategy: ship incremental updates more frequently instead of betting everything on blockbuster releases that set impossible expectations.