OpenAI has given GPT-6.1 Sol a much bigger working memory by default. The model now starts with a 500,000-token context length in Codex and ChatGPT Work, up from 272,000 tokens.
The change landed on October 11, 2026, less than two weeks after the model debuted.
What changed with GPT-6.1 Sol
OpenAI moved the default from 272,000 tokens to 500,000. That works out to an extra 228,000 tokens of room before users touch any settings.
The default is only the starting point. GPT-6.1 Sol supports a maximum context window of 1,050,000 tokens, and users can override the default to reach for more.
Advertisement
That ceiling splits into two parts. Up to 922,000 tokens can go to input, while as many as 128,000 tokens can go to output.
The surcharge goes away
The bigger default arrived alongside a pricing tweak. OpenAI eliminated long-context surcharges for subscription users of GPT-6.1 Sol.
For developers paying through the API, the published rates are $2 per million input tokens and $10 per million output tokens. Those prices are in line with what OpenAI charged for the model’s predecessor.
OpenAI framed the adjustment as a response to user feedback. The goal is to make the model better at handling long contexts and multi-step tasks.
Where Sol fits in the GPT-6 lineup
GPT-6.1 Sol launched on September 29, 2026, at OpenAI’s DevDay event. It belongs to the broader GPT-6 family of models.
OpenAI pitches Sol at people doing complex coding, heavy computer use and professional work. Its performance is described as closely aligned with the flagship GPT-6 Astra, while costing significantly less to run.
The model is available in ChatGPT Work and Codex. It targets Plus, Pro, Business, Enterprise and Edu users, with access through the OpenAI API as well.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.