OpenAI GPT-5.3-Codex: Here are the top features and performance upgrade

OpenAI has launched GPT-5.3-Codex, a new agentic coding model designed to assist professionals in a wide range of computer-based tasks. This model combines advanced coding abilities with strong reasoning and professional knowledge, allowing it to handle complex, long-running projects that involve research, tool usage, and detailed execution.

It can debug programs, deploy applications, monitor performance, write product requirements, create tests, conduct user research, edit content, and analyze data.

The model provides regular updates on progress and responds to user feedback in real time. It maintains context over extended sessions, making it feel like a collaborative partner in software development and other professional work.

The model brings strong performance on benchmarks. It scored 56.8% accuracy on SWE-Bench Pro, 77.3% on Terminal-Bench 2.0, and 64.7% on OSWorld-Verified. These results show improvements over earlier versions, with higher scores across coding, terminal operations, and computer-use tasks. It also performs strongly in areas like web development and cybersecurity challenges.

GPT-5.3-Codex (OpenAI)

GPT-5.3-Codex runs 25% faster than previous models due to optimized infrastructure and inference systems. This efficiency helps users complete tasks more quickly, especially in long sessions that require millions of tokens for iterative improvements.

The AI firm revealed that GPT-5.3-Codex also played a role in its own development. Early versions assisted the team with debugging training processes, managing deployments, and analyzing results. This approach sped up development and improved the model’s reliability for real-world use.

The model supports the full software development lifecycle, from initial coding and review to deployment and monitoring. It can operate autonomously on projects, making decisions and iterating based on goals. Integration is available through the Codex app, command-line interface, IDE extensions, and web platforms.

OpenAI has confirmed that access to GPT-5.3-Codex is provided to users with paid ChatGPT plans. It works across multiple tools, enabling seamless collaboration. Safety measures are in place, including specialized training and monitoring systems, with a focus on supporting defensive cybersecurity while addressing potential risks.

(source)