OpenAI has reduced the context window for its Codex model from 372k tokens to 272k tokens, a significant decrease in the amount of code and text the model can process in a single interaction.

This change affects developers using Codex for large codebase analysis, multi-file refactoring, and complex programming tasks that require extensive context. The 272k limit still supports substantial code reviews but may require more strategic context selection for very large projects.

The reduction likely reflects compute optimization or cost management decisions, though OpenAI has not publicly detailed the specific reasoning. Developers should adjust workflows accordingly, potentially splitting large tasks into smaller, focused sessions.