GPT-6.1 Sol 500K context in Codex, with no surcharge on subscription

The GPT-6.1 Sol 500K context is the change OpenAI made between October 11 and 12, 2026: the model now reads 500 thousand tokens by default in Codex and in ChatGPT Work, up from 272 thousand. It is still configurable up to 1 million and, when usage comes out of your subscription, there is no long-context surcharge. The API does not appear in the announcement.

What changed with the GPT-6.1 Sol 500K context in Codex

The default context is how much text the model can keep in mind at once without you touching any setting: files, instructions and the conversation history. With GPT-6.1 Sol that space went from 272 thousand to 500 thousand tokens, nearly double (500 ÷ 272 ≈ 1.84).

The news came out in the "28 days of shipping" series that OpenAI publishes in its developer forum, on day 7 of the series. It was shared by the Codex team and applies to Codex and ChatGPT Work.

The three concrete things

• 500 thousand tokens by default. Nothing to switch on: it is the new factory value.
• Up to 1 million, configurable. If your task needs it, the ceiling is still 1M.
• No long-context surcharge on subscription. When usage comes out of your subscription plan, no extra is charged for using a long context.

The fine print

The change was announced for subscription use. The API is not mentioned in the announcement, and secondary sources still list different rates above 272K tokens when you pay through the API. Before counting on it in a production workflow, check your plan and the current rate.

Also, do not confuse the default with the maximum: 500K is what you get without configuring anything, while 1M is the cap you can raise it to.

How to apply it in your business?

If you already use Codex on a subscription, the practical benefit is to stop splitting your repository or documentation into pieces just to make it fit. A wider context lets you hand over a whole module, several related files or a long contract in a single session, instead of writing summaries by hand.

A useful rule: start with the files that relate most to each other, review what the model answered, and only then widen the scope. More context does not replace good organization, and the question from the video still stands: which giant file would you feed it first?

To see how this model compares on price with its bigger sibling, read the full breakdown of GPT-6.1 Sol versus Astra.

Frequently asked questions

How many tokens does GPT-6.1 Sol read by default in Codex?

Now 500 thousand tokens by default, compared with 272 thousand before.

Can it reach 1 million tokens?

Yes, the context is still configurable up to 1 million.

Is there an extra charge for long context?

Not when usage comes out of your subscription. The API was not mentioned in the announcement.

Where does the change apply?

In Codex and in ChatGPT Work.

Conclusion

The GPT-6.1 Sol 500K context is a default-value change that helps anyone working with Codex on a subscription: nearly double the room and no long-context surcharge. The fine print is the API: it was not announced for it, so check your rate before assuming it.