OpenAI has released GPT-5.1-Codex-Max, a next-generation programming model designed for long-horizon reasoning, higher work efficiency and real-time interaction, the company said on Tuesday.

The model will replace GPT-5.1-Codex as the default option in the Codex integration interface.

The core upgrade is a new “compaction” mechanism, an architecture that allows the model to intelligently retain essential information and discard irrelevant details when approaching context-window limits. This enables sustained performance across millions of tokens.

In internal tests, GPT-5.1-Codex-Max completed complex tasks running for more than 24 hours, including multi-step code refactoring and autonomous debugging. OpenAI said the mechanism improves token efficiency by about 30%, reducing development costs and latency.