GPT-5.5-Cyber

// Overview

GPT-5.5-Cyber is OpenAI's most permissive cybersecurity tier, opened in limited preview on May 7, 2026 as part of the Trusted Access for Cyber (TAC) program. It is built on the same underlying GPT-5.5 model that became OpenAI's flagship in late April 2026, but is tuned to issue fewer classifier-based refusals on authorized security tasks for a small set of vetted, invite-only defenders and approved partner organizations.

The model accepts text and image input and shares GPT-5.5's core specs: a roughly 1.05M-token context window, up to 128K output tokens, and a December 2025 knowledge cutoff. OpenAI is explicit that GPT-5.5-Cyber is 'not intended to deliver a major jump in raw cyber performance' over standard GPT-5.5 — the difference is permissiveness on dual-use security workflows, not raw capability. For most defenders, OpenAI recommends standard GPT-5.5 with Trusted Access for Cyber rather than the Cyber tier.

Access is gated. Individuals apply at chatgpt.com/cyber and enterprises via OpenAI's enterprise TAC form; the Cyber tier is the highest, invite-only level. Beginning June 1, 2026, individual members accessing OpenAI's most permissive cyber models must enable Advanced Account Security with phishing-resistant protections, or organizations can attest to phishing-resistant single sign-on. Safeguards still block malicious activity such as credential theft, stealth/persistence techniques, malware deployment, and exploitation of third-party systems.

// Benchmarks

1. Expert-level cyber tasks (UK AISI advanced suite, avg pass rate)71.4%

Scores on a 0–100 scale (25-point gridlines); higher is better. Each benchmark links to its published source.

// Pricing

OpenAI does not publish a separate price for the invite-only GPT-5.5-Cyber tier; these are the underlying GPT-5.5 API token rates. Access to the Cyber tier is granted through the Trusted Access for Cyber program, not the public price list.

Pricing source ↗

// Strengths

• Fewer automated refusals on authorized, dual-use security workflows for vetted defenders
• Vulnerability identification and triage, patch validation, and detection engineering
• Malware analysis and binary reverse engineering
• Authorized penetration testing, red teaming, and exploit proof-of-concept validation for coordinated disclosure
• Inherits GPT-5.5's frontier agentic reasoning, ~1.05M-token context, and tool use

// Best for

• Reach for it when a vetted security team needs an authorized model that won't over-refuse on legitimate offensive testing (pentest, red teaming).
• Reach for it for high-severity vulnerability research, exploit validation, and patch validation under coordinated disclosure.
• Reach for it for malware analysis and binary reverse engineering inside an approved, monitored environment.
• Use standard GPT-5.5 with Trusted Access for Cyber instead for the vast majority of defensive workflows that do not need the most permissive tier.

// How to access