According to a report by The Information, GPT-5.4 may feature a context window of up to one million tokens, more than double the 400,000 tokens supported by the current GPT-5.2 model. With this upgrade, OpenAI would reach parity with competing models from Google and Anthropic that already offer extended context capabilities.
The new model is also expected to perform more reliably on long-running tasks that can last several hours, reducing the likelihood of errors during complex operations. This improvement could be particularly important for OpenAI’s Codex programming tool and other developer-focused applications.
Another notable feature reportedly under development is an “extreme” reasoning mode. In this mode, the model would allocate significantly more computing time to difficult questions, allowing for deeper analysis and more precise answers. The feature is aimed primarily at research applications rather than everyday users who prioritize faster responses.
OpenAI is also reportedly planning more frequent model updates, sometimes on a monthly basis. According to The Information, this strategy is intended to prevent unrealistic expectations from building up, as happened ahead of the GPT-5 launch. In recent months, OpenAI’s user growth has reportedly been slower than internal forecasts.
ES 
EN