OpenAI has recently unveiled the latest version of its cutting-edge AI model, GPT-5.2, along with an optimized coding model called Codex. The new Codex model is specifically designed for agentic building and is touted as the most advanced agentic coding model yet for complex, real-world software engineering. This update brings more agentic use cases to OpenAI’s arsenal, offering stronger cybersecurity capabilities and optimized for long-horizon work with agents.
According to OpenAI’s blog post, GPT-5.2-Codex represents a significant step forward in how advanced AI can support real-world software engineering and specialized domains like cybersecurity. The company is rolling out access to the new Codex model for paid ChatGPT users and is working towards enabling access for API users in the coming weeks. Additionally, OpenAI is piloting a program with invite-only trusted users to access more permissive models for defensive cybersecurity work, striking a balance between accessibility and safety.
The advancements in cybersecurity capabilities with the GPT-5.2-Codex model are impressive. OpenAI tested the model on various benchmarks, including Capture-the-Flag (CTF) evals, CVE-Bench, and Cyber Range. The model excelled in CTF evals, scoring 87%, and outperformed other models in CVE-Bench with a score of 87%. In the long-form Cyber Range test, the model had a combined pass rate of 72.7%, showcasing its robust cybersecurity capabilities.
OpenAI is also focused on a cybersecurity deployment project to ensure the safe and responsible use of its frontier AI models. The company aims to provide trusted defenders with access to frontier AI cyber capabilities to accelerate cyberdefense efforts. With the introduction of GPT-5.2-Codex and potential future models, OpenAI is committed to balancing the deployment of advanced AI models with the necessary tools for defensive cybersecurity.
The new Codex model is expected to enhance long-horizon work through compaction, offering strong performance on extensive code changes. It also features improved performance on Windows and has shown improved accuracy in benchmark testing. Since its preview launch in May, Codex has gained traction in the enterprise AI builder space, enabling agentic and vibe coding for users. Alongside other coding agents like Windsurf, Cursor, and Claude Code, Codex is changing the landscape of AI-powered coding platforms, moving beyond simple code completion to generating and initiating complex coding projects for users.
Overall, OpenAI’s latest release of GPT-5.2 and Codex represents a significant leap forward in the field of AI-powered coding and cybersecurity. The company’s focus on responsible deployment and enhancing defensive cybersecurity capabilities ensures that these advanced AI models can be used effectively and safely in real-world applications.
