OpenAI has raised the default context window for GPT-6.1 Sol to 500,000 tokens in Codex and ChatGPT Work, and dropped the long-context surcharge for subscribers. More work now fits in a single session.
GPT-6.1 Sol's Default Context Jumps to 500K Tokens
OpenAI lifted the default context length for GPT-6.1 Sol from roughly 272,000 tokens to 500,000 tokens when the model is used through Codex and ChatGPT Work. Tibo Sottiaux, who leads OpenAI's Codex team, announced the change on X on October 12. That's an increase of about 80%.
Context window is the amount of text a model can hold in mind at once, including your instructions, any files you paste in, and the conversation so far. When you run past it, the model starts forgetting the earliest parts of the session. A bigger default window means fewer interruptions on long jobs.
The number matters most for people who feed the model big codebases, long documents, or multi-hour working sessions. Those are exactly the tasks Codex and ChatGPT Work are built around.
The Long-Context Surcharge Is Gone for Subscribers
The other half of the announcement is the pricing change. Under OpenAI's previous setup, prompts that ran past the short-context threshold, defined as 272,000 input tokens, were billed at a higher rate. The company charged roughly double for input and cached input, and 1.5 times for output, on those oversized requests.
For subscribers using Codex and ChatGPT Work, that extra charge is now gone. If your work regularly pushes past the old 272K line, the shift could cut what you pay for the same task.
What improved
- Default context doubles from about 272K to 500K tokens
- No long-context surcharge for Codex and ChatGPT Work subscribers
- Longer sessions before the model starts dropping earlier content
What didn't change
- The model's full ceiling is still a separate, higher number
- Pricing beyond the default window still depends on the published rates
- The change targets subscriptions, not pay-as-you-go API billing
What 500K Tokens Actually Fits
To put the window in perspective: 500,000 tokens is a lot of text. It's roughly the size of a handful of full-length books, or a large slice of a code repository.
So what does that buy you when you're actually working?
In practice, the upgrade means you can hand the model a long chat history plus a stack of files and still have room to keep working. Before, hitting the ceiling forced a choice: trim the input, start a fresh session, or pay more to keep going.
For developers, the useful part is continuity. A working session that used to spill into a second window can now stay in one. That cuts the awkward work of re-explaining context the model already had. It also means fewer resets. Less busywork.
Where the Numbers Come From
The 272K to 500K figures and the surcharge removal come from OpenAI's own X post and its pricing page. Third-party trackers list GPT-6.1 Sol's full context ceiling at about 1.05 million tokens, with up to 128,000 output tokens, and standard API pricing around $2 per million input tokens, $0.10 per million cached, and $10 per million output.
Those fuller specs come from outside aggregators, so treat them as a guide rather than an official statement. The core change here, the new default and the dropped fee, is OpenAI's own.
What This Means for Your Workflow
If you use Codex or ChatGPT Work on subscription, the practical win is fewer mid-task resets. Long debugging sessions, whole-project reviews, and large document passes all benefit. Fewer resets, less friction.
Watch OpenAI's pricing page for the finer print on what counts as "long" once the new default settles in. And if you pay per token through the API instead of a subscription, check whether the surcharge change reaches you too, since the announcement focuses on the subscription side.
