OpenAI has stopped charging a long-context fee for GPT-6.1 Sol when the usage comes out of a ChatGPT subscription. Tibo (@thsottiaux), who leads Codex and ChatGPT Work, posted the follow-up at 07:48 Beijing time on October 12: "With this change, we are not charging long context fees for GPT-6.1 Sol when drawing usage from the subscription." That line is his wording, and it applies to subscription billing — not to the API.
Minutes earlier, at 07:45, he posted it as day 7 of the Codex plan: the default context length for GPT-6.1 Sol in Codex and Work had moved from 272k to 500k tokens, roughly an 80% increase. The longer default, he wrote, should cut down on how often the system has to compact a long thread, and it should hold more context and nuance across longer tasks. He called 500k the new sweet spot for most users after careful testing, with room to go higher later. Users can still override the setting higher or lower; the model's maximum remains 1M tokens.
The API side is different. OpenAI's GPT-6.1 Sol model page still says prompts with more than 272K input tokens are priced at 2x input and cache rates and 1.5x output for the full request. So the fee drop is a subscription change in Codex and Work. If you are calling the model through the API, the long-context multiplier above 272K is still there.
How much included usage each ChatGPT plan actually gets for Codex is still opaque in places. Our guide "OpenAI Keeps Codex's Weekly Caps Vague: What Each ChatGPT Plan Actually Gets" covers what OpenAI does and doesn't publish about those caps.
