Skip to main content

OpenAI Doubles GPT-6.1 Sol's Memory to 500K Tokens, No Extra Cost

OpenAI Expands GPT-6.1 Sol's Context Window to 500,000 Tokens

If you've ever been deep in a coding session, only to have your AI assistant suddenly forget what you were working on, you're not alone. OpenAI is tackling that headache head-on. On October 12, Thibault Sottiaux from OpenAI announced that the default context window for GPT-6.1 Sol in Codex and Work environments has been bumped from 272,000 to 500,000 tokens—an 80% jump in what you might call the model's "short-term memory."

Why This Matters for Heavy Users

A bigger context window means the model can hold onto more information before it has to start summarizing or discarding details. That awkward moment when your AI hastily compresses context—losing nuance and important specifics—will become much rarer. For tasks that require juggling lots of details over a long stretch, this is a game-changer. You won't have to constantly clean up or repeat yourself.

Thibault noted that 500,000 tokens hits the "sweet spot" between performance and experience for most users. It's been thoroughly tested, so you're not sacrificing speed for memory.

You're Still in Control

Of course, if you need even more room, GPT-6.1 Sol supports up to 1,000,000 tokens. Whether you want to max it out or dial it back, the choice is yours. No one-size-fits-all here.

And here's the kicker: long texts won't be charged separately. So you can stretch those tokens without watching your bill spike.

What This Means for Developers

For developers working on large codebases, lengthy documents, or multi-step workflows, this update could mean fewer interruptions and more seamless interactions. The model can keep more of your project in mind, reducing the need for manual context management. It's a subtle but significant shift that could make AI feel less like a tool and more like a collaborator.

Key Points:

  • Default context window for GPT-6.1 Sol in Codex and Work environments increased from 272,000 to 500,000 tokens.
  • 80% boost in "short-term memory" reduces frequent context compression.
  • 500,000 tokens is the optimal balance for most users, tested for performance.
  • Users can manually set context up to 1,000,000 tokens.
  • No extra charge for long texts—cost stays predictable.