# GPT-6.1 Sol Gets 500K Default Context in Codex and Work

OpenAI raises GPT-6.1 Sol's default context from 272K to 500K in Codex and Work and removes long-context fees for usage drawn from a subscription.

By Paras Madan | 2026-10-12


OpenAI has increased the default context length for GPT-6.1 Sol in Codex and ChatGPT Work from 272,000 to 500,000 tokens. Alongside the change, it is removing long-context fees when the model draws usage from a subscription.

## What the 500K announcement covers

OpenAI's Tibo announced the increase late on October 11 UTC, as day seven of the team's shipping series. His [announcement](https://x.com/thsottiaux/status/2109430343346581527) specifically covers GPT-6.1 Sol used through Codex and Work. Moving from 272K to 500K adds 228,000 tokens, an increase of roughly 84%.

Tibo says the setting was tested to reduce the frequency of compaction and help tasks that need to retain context and nuance over time. Users can still override the default in either direction, with a stated maximum of 1M for these products.

An agent editing related modules may benefit from retaining earlier requirements and test results. Whether the larger window improves its work on a particular codebase still needs testing.

## Subscription usage and API billing have different rules

In a [follow-up clarification](https://x.com/thsottiaux/status/2109430894604026039), Tibo said OpenAI is no longer charging long-context fees for GPT-6.1 Sol when usage is drawn from the subscription. That scope matters for anyone switching between a ChatGPT-backed coding session and an application billed through an API key.

The current [GPT-6.1 Sol API documentation](https://developers.openai.com/api/docs/models/gpt-6.1-sol) still specifies a surcharge for prompts above 272K input tokens: input and cache rates are doubled, while output is charged at 1.5 times the normal rate for the full request. The subscription announcement therefore should not be applied to an API cost estimate.

A context window also measures something different from a usage allowance. It describes how much material the model can handle in a request. Raising the default does not give a subscriber 84% more total usage, and the fee clarification does not make long sessions unlimited.

## What to check on your next long task

If you previously changed your context settings, review that override before comparing results with the new default. A manually chosen value can make a comparison misleading.

For a useful test, repeat a task with a clear acceptance check. Watch when compaction happens and whether the agent still follows a requirement introduced early in the conversation. Compare the final result and usage as well as the size of the context window. A larger number in the interface is only one part of that assessment.

For teams documenting AI coding workflows, the billing detail belongs next to the setup instructions. State whether a workflow uses subscription-backed access or API billing so readers can apply the right cost assumptions before running a large job.

I'm Paras Madan. I work across AI engineering and distribution, helping technical products reach the developers who can use them.

I help teams explain changes like this through clear technical writing, including release explainers and developer guides that make product limits and billing rules easy to follow.

[Book a 30-minute call](https://cal.com/paras-madan/30min)