OpenAI Launches GPT-5.1-Codex-Max for Advanced Long-Horizon Coding Tasks

OpenAI has unveiled GPT-5.1-Codex-Max, a groundbreaking evolution in AI-assisted software engineering. Built on the foundation of GPT-5.1, this model specializes in managing highly complex, multi-stage development workflows that previously posed insurmountable challenges for traditional models. Its key innovation is a process called compaction, which enables the model to coherently work across millions of tokens, supporting prolonged, autonomous coding sessions lasting over 24 hours.

This new model significantly enhances AI’s role in software development by optimizing token efficiency—reducing the number of "thinking tokens" by approximately 30% without compromising performance. It can undertake extensive refactors, deep debugging, and iterative build processes with minimal human intervention, making it an invaluable tool for complex projects at an enterprise level. For example, developers can now achieve full-stack feature implementations or aggressive security audits in one sustained session, with the AI maintaining context and coherence through the compaction technology.

Furthermore, GPT-5.1-Codex-Max offers an Extra High (xhigh) reasoning mode, aimed at the most demanding problem-solving scenarios that require intense computation and analysis. This mode extends the model’s capabilities, enabling it to consistently address sophisticated tasks such as multi-layered architecture design, cybersecurity threat hunting, and algorithmic research, with notably improved accuracy and depth.

Access to GPT-5.1-Codex-Max is available through API, IDE integrations, and cloud platforms. Its internal evaluation benchmarks report a 77.9% accuracy on SWE-bench verified tasks, solidifying its position as a leading AI coding assistant. It now surpasses previous models in both effectiveness and cost efficiency, facilitating faster, smarter development cycles for overburdened engineering teams.

The deployment of GPT-5.1-Codex-Max marks a new frontier in AI-driven software development, with its ability to sustain autonomous, long-duration workflows representing a leap toward fully agentic programming assistants. As OpenAI continues to refine these capabilities, future iterations are expected to further deepen AI integration within software engineering, ultimately transforming how tech companies approach large-scale coding tasks.