OpenAI Announced the Autonomous Coding Model GPT-5.2 Codex

New Standard in Autonomous Coding: GPT-5.2 Codex

OpenAI introduced the new Codex model built on the GPT-5.2 architecture to accelerate software development processes. Compared with previous generations, the system has more advanced autonomous capabilities and promises high performance especially in long-term context management and complex system setups.

The model’s most notable feature is its built-in context compression technology. Thanks to this, while analyzing code directories of thousands of lines, the project’s overall structure and memory can be preserved. In addition, with its visual understanding capability, it can directly turn technical diagrams and interface designs (UI) into working code.

The standout technical capabilities and performance data of GPT-5.2 Codex are as follows:

• Advanced Terminal Use: The model, optimized for Windows environments, can autonomously perform system administration tasks such as code compilation, server setup, and model training.
• Benchmark Scores: A success rate of 56.4% was achieved in the SWE-Bench Pro test, and 64% in the Terminal-Bench 2.0 test.
• Cybersecurity Analysis: The model, with enhanced defensive capabilities, proved its effectiveness in code auditing by identifying three different security vulnerabilities in the React library.

On the accessibility side, OpenAI initially began offering GPT-5.2 Codex to paid ChatGPT users through Codex CLI and IDE plugins. For cybersecurity experts, the “Trusted Access” program was launched.