Engineer Tibo, who is responsible for Codex and ChatGPT at OpenAI, recently announced on the X platform that the 1 million Token context window of the GPT-5.6 Sol model in Codex has been expanded from being limited to API keys to also being available in ChatGPT account scenarios. This means that the restriction of cutting off history when dealing with large projects for subscribed users has finally been lifted.
Subscribed users can manually enable the million-token window
Tibo clearly stated that GPT-5.6 Sol in Codex supports a context length of 1 million tokens, which was previously only available for API keys. Now, "the switch has just been turned on," and it has already taken effect through ChatGPT accounts. Users with Plus, Pro, and other subscription plans can manually configure and enable the full context window of nearly 1.05 million tokens when using the Codex programming agent.
A larger context window allows Codex to retain more code, tool outputs, and conversation records before automatic compression, especially suitable for professional scenarios such as large-scale codebase refactoring, complex debugging, and long sessions. For developers maintaining production-level repositories, this adjustment directly alleviates the previous pain point of frequently "not enough" context.
Default length is carefully considered, enabling requires caution
Tibo also emphasized that the current default context length is the result of comprehensive optimization of performance and cost, not set arbitrarily. He reminded users, "You can decide for yourself." Several developers also pointed out that combined with Codex's server-side compression mechanism, the default configuration is sufficient for most daily tasks.
However, some users have reported that after enabling the million-Token context, the subscription quota is consumed significantly faster, with a single intensive use possibly consuming a large portion of the weekly quota. It is recommended to enable it only when necessary, rather than as the default setting for daily use, to avoid reaching the usage limit prematurely on critical projects.
