OpenAI has released GPT-5.3-Codex, an advanced coding model that enhances functionality beyond basic code assistance. This upgrade significantly boosts performance, running 25% faster than its predecessor, GPT-5.2-Codex, and is designed for interactive, multi-step tasks, such as debugging and managing code workflows. Improved reasoning capabilities make it adept at handling complex coding challenges, achieving high scores on evaluation benchmarks, including SWE-Bench Pro and Terminal-Bench 2.0.
GPT-5.3-Codex has been used in its own development for tasks like debugging and deployment management, showcasing its self-improvement potential. Additionally, it is classified as a “High capability” model for cybersecurity, directly trained to identify software vulnerabilities. OpenAI is also piloting Trusted Access for Cyber to enhance defensive security research. Users can access GPT-5.3-Codex through paid ChatGPT plans, with API access expected soon, leveraging cutting-edge NVIDIA infrastructure for optimal performance.
Source link
