OpenAI Releases GPT‑5.6‑Cyber Hacking Model Amid Regulatory Scrutiny and Price Cuts

OpenAI unveiled GPT‑5.6‑Cyber, described as its first “offense‑grade” hacking model, designed to probe software and network security flaws in a controlled research environment, marking a new direction for AI‑driven cybersecurity testing.

GPT‑5.6‑Cyber joins the broader GPT‑5.6 suite that also includes Sol, Terra and Luna variants, which are currently limited to a small group of preview partners cleared by the U.S. government amid ongoing AI regulatory discussions.

Ultrafast Mode added to GPT‑5.6‑Sol claims a 14‑times speed boost, enabling developers to run large prompts, code generation and data‑analysis tasks far quicker than earlier models, while OpenAI also opened unlimited free chat access for non‑paying users.

Price Cuts announced for two GPT‑5.6 models as enterprises voice cost concerns; OpenAI reduced subscription fees and promised broader availability after a government‑requested delay, aiming to keep advanced AI tools affordable for business customers.