Overview
- OpenAI began rolling out GPT-5.5 on Thursday to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex, with a Pro tier live and API access announced with pricing.
- The model is framed as an agent that plans steps, uses tools, checks its work, and completes multi-part computer tasks with far less prompting.
- Performance rose on real-world tests, including 82.7% on Terminal-Bench 2.0 for command-line workflows, while matching GPT-5.4’s per-token speed and using fewer tokens for many coding tasks.
- OpenAI says this release includes its strongest safeguards yet, with preparedness evaluations, external red teaming, a bio bug bounty, and a “High” risk label for cyber and biological capabilities.
- Early enterprise testers such as Bank of New York report higher response quality and fewer hallucinations, which could speed deployment in regulated workflows.