OpenAI has taken a new step in the availability of its most advanced programming model by making GPT-5.2 Codex accessible via its API.
This means the model, which has been available in Codex for some time, is now also directly accessible to developers who want to integrate it into their own applications and tools. This is according to Neowin.
The introduction of GPT-5.2-Codex follows the broader release of the GPT-5.2 model family, which, according to OpenAI, delivers clear improvements in intelligence and reliability. Within that series, GPT-5.2-Codex focuses specifically on software development. The model is designed for agentic workflows in which AI can independently perform long-term and complex programming tasks, such as building new functionality, restructuring existing code, and detecting errors in large codebases.
With its availability via the Responses API, GPT-5.2-Codex is effectively on par with the version used in Codex itself. Developers can now access the model directly from their own development environments and applications. According to OpenAI, the main focus has been on improving performance over longer periods, including through more efficient handling of context, and on increasing reliability in large-scale projects.
Generating code without human intervention
Support for GPT-5.2-Codex is now visible in various popular development tools and IDEs. The companies behind these tools say the model excels at long-running tasks that previous models struggled to maintain consistency. In practical tests, the model was used for extremely large projects, generating and managing code autonomously for days without human intervention.
The move to make GPT-5.2-Codex available via an API underscores OpenAI’s focus on developers and professional applications. Whereas the company positioned earlier generations as assistants for individual programming tasks, the role of this type of model is increasingly shifting toward full-fledged software agents.