OpenAI says GPT-5.2-Codex is its ‘most advanced agentic coding model yet’ – here’s what developers and cyber teams can expect

OpenAI has announced a big update for its Codex coding agent as the company targets big improvements for developers and cybersecurity teams.

According to the company, GPT-5.2-Codex, a version of GPT-5.2 optimized specifically for Codex, is its most “advanced agentic coding model yet”.

Improvements touted by OpenAI include more efficient “long-horizon work” and better performance working on large code changes, such as refactoring and migrations.

Improved performance in Windows environments were also revealed alongside stronger cybersecurity capabilities.

“GPT‑5.2-Codex builds on GPT‑5.2’s strengths⁠ in professional knowledge work and GPT‑5.1-Codex-Max⁠’s frontier agentic coding and terminal-using capabilities,” the company said in a blog post.

“GPT‑5.2-Codex is now better at long-context understanding, reliable tool calling, improved factuality, and native compaction, making it a more dependable partner for long running coding tasks, while remaining token-efficient in its reasoning.”

Codex first launched in May this year, and since the launch and integration of GPT-5, uptake of the coding agent has skyrocketed, according to OpenAI. Indeed, a host of large organizations are now using the tool to supercharge software engineering activities.

“Codex usage has more than 20x’d since the launch of GPT-5-Codex in August, and customers like Cisco, Ramp, Virgin Atlantic, Vanta, Duolingo, and Gap are using the agent to accelerate their engineering teams.”

GPT-5.2-Codex performance improvements

According to OpenAI, GPT-5.2-Codex boasts “state-of-the-art performance” on SWE-Bench Pro and Terminal-Bench 2.0.

These are benchmarks aimed at testing agentic AI performance across an array of realistic environments. In terms of accuracy, the new model recorded an SWE Bench Pro mark of 56.4%, marking marginal improvements on GPT-5.2 (55.6%) and GPT-5.1 (50.8%).

Meanwhile, on Terminal-Bench 2.0 the model recorded an accuracy rating of 64.0% compared to GPT-5.2’s 62.2% and GPT-5.1-Codex-Max’s 58.1%.

“With these improvements, Codex is more capable at working in large repositories over extended sessions with full context intact,” the company explained.

“It can more reliably complete complex tasks like large refactors, code migrations, and feature builds — continuing to iterate without losing track, even when plans change or attempts fail.”

Cyber improvements

On the cybersecurity front, OpenAI said the new model marks another significant jump in terms of capabilities.

Initial testing of the model in capture-the-flag challenges shows it scored a higher accuracy rating than GPT-5.2, GPT-5-Codex, and GPT-5.1-Codex-Max, for example.

“When charting performance on one of our core cybersecurity evaluations over time, we see a sharp jump in capability starting with GPT‑5-Codex, another large jump with GPT‑5.1-Codex-Max and now a third jump with GPT‑5.2-Codex,” OpenAI noted.

“We expect that upcoming AI models will continue on this trajectory.”

Notably, the company said it plans to introduce new safeguards for the model to prevent potential misuse.

Earlier this month, ITPro reported that OpenAI plans to collaborate with red teaming organizations to test systems, and will set up a ‘trusted partner system’ for future model experimentation by security researchers.

The company is now inviting security researchers to participate in an upcoming pilot for the Cyber Trusted Access scheme.

How to get your hands on GPT-5.2-Codex

GPT-5.2 Codex is available now for all paid ChatGPT users, OpenAI confirmed. Similarly, the company said it’s “working towards safely enabling access” for API users in the coming weeks.

“In parallel, we’re piloting invite-only trusted access to upcoming capabilities and more permissive models for vetted professionals and organizations focused on defensive cybersecurity work,” the company added.

“We believe that this approach to deployment will balance accessibility with safety.”

FOLLOW US ON SOCIAL MEDIA

Make sure to follow ITPro on Google News to keep tabs on all our latest news, analysis, and reviews.

You can also follow ITPro on LinkedIn, X, Facebook, and BlueSky.