Codex 28-day program, Day 7: GPT-6.1 Sol's default context jumps to 500k in Codex and Work

#What changed

Tibo (@thsottiaux) posted Day 7 of the Codex program on Oct 11: the default context window for GPT-6.1 Sol in Codex and Work went from 272k to 500k tokens, roughly 80% more. He says the new default was tested as the sweet spot for most users and should mean fewer compactions on long tasks that need to keep a lot of detail around. The model's maximum is still 1M tokens, and you can still set the window higher or lower yourself. In a follow-up he added that there are no long-context fees for GPT-6.1 Sol when usage is drawn from a subscription. This replaces the smaller default Tibo described on Oct 10.

#Who is affected

Anyone running GPT-6.1 Sol through Codex or Work on a ChatGPT subscription. API pricing wasn't mentioned.

#What to do now

If you raised model_context_window by hand to avoid compactions, try dropping that override and using the new default. If you lowered it to stretch your usage, keep your override, since long threads still fill the window. For the first few days, keep an eye on your usage meter during long sessions.

#What is not confirmed

This comes from Tibo's posts on X; we haven't seen release notes or docs updated yet. He didn't say whether a bigger default changes how quickly long sessions use up plan limits, beyond dropping the long-context fee.