![](\"https://images.ctfassets.net/jdtwqhzvc2n1/1vFvZadzVgpBpKJvEUCDS9/e9b912eb53c0fda59b01e8676b3613b3/ChatGPT_Image_Aug_10__2026__07_36_04_PM.png?w=300&q=30\")
OpenAI Launches GPT-5.6-Cyber for Advanced Cybersecurity Tasks
Today, OpenAI introduced GPT-5.6-Cyber, a specialized model tailored for advanced vulnerability research and exploit development for authorized defenders. This model is designed to handle tasks that general-purpose models often refuse.
GPT-5.6-Cyber is a fine-tuned version of OpenAI’s GPT-5.6 Sol, the most advanced general model unveiled in June. It is specifically trained to excel in advanced cybersecurity tasks such as identifying zero-day vulnerabilities and developing exploit chains.
OpenAI has enhanced GPT-5.6-Cyber to reduce refusals on high-risk cybersecurity requests, including those that could be used for both legitimate defensive and malicious offensive purposes. In an internal benchmark test, GPT-5.6-Cyber achieved a completion rate of 95%, outperforming its predecessor GPT-5.5-Cyber at 57.3%, and significantly surpassing GPT-5.6 Sol’s completion rate of just 1.5%.
According to OpenAI researcher Eric Wallace, GPT-5.6-Cyber represents OpenAI’s first major effort to enhance capabilities for advanced cybersecurity tasks like exploit development.
Pricing and Access
GPT-5.6-Cyber is not available to all ChatGPT or API customers. Organizations must be accepted into OpenAI’s Daybreak cybersecurity program, specifically the Daybreak Red tier, to gain access to dedicated cybersecurity models like GPT-5.6-Cyber.
Another tier, Daybreak Blue, provides a broader range of enterprises with access to general models like GPT-5.6 Sol, with some guardrails lifted to accommodate more cybersecurity applications.
OpenAI has priced GPT-5.6-Cyber at $12.50 per million input tokens and $75 per million output tokens, with cached input tokens costing $1.25 per million. Comparatively, GPT-5.6 Sol is listed at $5 per million input tokens and $30 per million output tokens for short-context use.
Access to GPT-5.6-Cyber requires separate approval and provisioning through Daybreak Red.
Daybreak Red vs. Daybreak Blue
Daybreak Red is designed for approved security teams engaged in advanced, authorized cybersecurity work, such as vulnerability research and exploit validation. Organizations applying for access must demonstrate a clear professional need for the model and adhere to strict security controls.
For enterprises that do not qualify for Daybreak Red or do not require specialized access, OpenAI recommends Daybreak Blue. This tier offers approved defenders access to general-purpose models like GPT-5.6 Sol, tailored for legitimate defensive applications.
Daybreak presents two pathways for enterprises to leverage OpenAI’s cybersecurity capabilities, with Daybreak Blue serving as a starting point for organizations seeking AI support for everyday security tasks.
… and so on.
