OpenAI Launches GPT-5.6-Cyber Hacking Model With Ultrafast 14x Speed Boost

OpenAI announced the release of GPT-5.6-Cyber, its first model described as “offense‑grade” and built for penetration testing and cybersecurity research, marking a notable shift toward AI tools that can assist security professionals in identifying system weaknesses.

GPT-5.6-Cyber can generate exploit code, write vulnerability‑scanning scripts, and simulate attack vectors, but OpenAI says access will be limited to vetted partners, with strict usage policies and monitoring to prevent malicious misuse.

Ultrafast Mode rolled out with GPT-5.6, claims a 14‑times speed boost over the standard Sol configuration, allowing the model to process up to 28,000 tokens in three minutes, which could accelerate real‑time threat analysis for security teams.

OpenAI will retire GPT-5.5 on October 14, urging developers to migrate to the newer GPT-5.6 platform, update their APIs, and review the revised compliance guidelines before the shutdown to avoid service interruptions.