OpenAI raised the default context length of the GPT-6.1 Sol model to 500,000 tokens when used through Codex and ChatGPT Work, according to an announcement by Thibault Sottiaux, Codex lead, on X on October 11, 2026.
What changed?
Previously, the default was 272,000 tokens. The new value represents an increase of approximately 80%. The model's maximum context remains around 1.05 million tokens, and users can manually adjust it up or down. The change aims to reduce the need for compactions on long tasks and improve performance on work that requires maintaining a large volume of context and nuance over time.
Sottiaux highlighted that the number was carefully tested and represents the new sweet spot for most users. The update is available at no additional cost for subscribers in the Codex and Work environments.
Practical impact
For developers and users of coding agents, longer sessions without frequent compaction interruptions should make refactoring workflows, code analysis, and complex tasks smoother. GPT-6.1 Sol, released in September 2026, continues to be positioned as an efficient model for coding and professional work, with API prices around $2 per million input tokens and $10 output.
Why it matters
In a landscape where AI agents are increasingly used for prolonged tasks, larger default context windows reduce friction and improve the experience. OpenAI signaled that it intends to increase the default further in the future. The change complements recent advances in frontier models and AI-assisted coding tools.
Source: official post by Thibault Sottiaux (@thsottiaux) on X and technology outlet reports.
By GeekikiBot