OpenAI unveils major GPT‑5.5 upgrades for cybersecurity and chat performance

OpenAI has introduced GPT‑5.5 Cyber, a specialised variant of its latest model aimed at authorised cybersecurity workflows. The new system assists defence teams in identifying, validating and remediating software vulnerabilities, offering multi‑step reasoning, code‑review assistance and automated patch suggestions. Access is restricted through a Trusted Access for Cyber framework that requires identity verification, account controls and monitored environments.

The company also rolled out an update to GPT‑5.5 Instant, the default model behind ChatGPT for all users. The upgrade improves the model’s ability to retain conversation context, understand user intent and handle complex, multi‑constraint requests. OpenAI reports a 52.5 % drop in hallucinations and a 37.3 % reduction in factual errors compared with the previous version. New features include better geographic awareness for local recommendations, pronunciation guidance in over 60 languages, UI enhancements such as pinned conversations and project grouping, and tighter privacy controls for third‑party app permissions.

In parallel, OpenAI has added more than 30 new models, tools and features to its API in the past six months, including GPT‑5.5 Pro, mini and nano variants designed for cost‑effective high‑volume workloads, and an open‑source Privacy Filter for local PII masking. Pricing for the API is set at $5 per million input tokens and $30 per million output tokens. These rapid releases expand OpenAI’s ecosystem while raising both opportunities and integration challenges for developers.