OpenAI ships GPT-6 Cyber for Daybreak Red
SummaryNov 5, 2026: OpenAI introduced GPT-6 Cyber, a cybersecurity model for finding and validating exploitable vulnerabilities inside researcher-set scope. OpenAI claims 48.5% on ExploitGym vs 32.1% for GPT-5.6 Cyber, Preparedness Critical, and select Daybreak Red API + Codex access only.
· MODEL Editors · 1 min read
Some links may be affiliate links. Disclosure
OpenAI just introduced GPT-6 Cyber (Nov 5, 2026): a cybersecurity model trained to find and validate exploitable vulnerabilities while staying inside the scope researchers set. Pitch is more capability than GPT-5.6 Cyber, plus tighter alignment on authorized bounds. Primary: OpenAI's post (opens in a new tab).
What OpenAI claims. On ExploitGym (exploit a specified vulnerability), GPT-6 Cyber scores 48.5% vs 32.1% for GPT-5.6 Cyber — the ~51% relative lift OpenAI cites. On ExploitGym Honeypot (stay in scope when easy out-of-scope services sit nearby), it avoided out-of-scope exploits in 99.93% of tests without the deployment prompt and 100% with it, vs 94.49% for GPT-5.6 Cyber without. Preparedness Framework rating: Critical cybersecurity (up from High for GPT-5.6 Cyber).
Real-world use OpenAI reports. Researchers used it for 948 Linux privilege-escalation exploits and 81 previously unknown high-severity Firefox vulnerabilities (Mozilla releasing fixes). Named customers / partners in the post include Mozilla/Firefox, Cisco, Calif, CodeRabbit, Palo Alto Networks, Cognition, SentinelOne, and Perplexity. The @OpenAI X post highlighted Mozilla, Palo Alto Networks, Calif, and Cisco.
Who it's for. Select approved Daybreak Red customers via API + Codex at launch; OpenAI says access expands over time, and GPT-5.6 Cyber is going to all Daybreak Red. Not a public ChatGPT feature for everyone.
Honest caveats. Access is gated (select Daybreak Red). The Critical rating is OpenAI's own Preparedness call. Benchmark and real-world counts above are OpenAI claims. MODEL has not tested GPT-6 Cyber.
Disclosure: Not sponsored. No affiliate links. Not investment advice.
MODEL Editors
Sources
Keep reading
Tool Update
ChatGPT Images 2.6 adds GIFs and faster API models
OpenAI shipped Images 2.6 in ChatGPT, Codex, and the API on Nov 4: native GIFs, saved reference images, and GPT-Image-2.6 Flare + Sunburst.
· MODEL Editors
Tool Update
Kimi K3.1 Pro open weights are out on Hugging Face
Moonshot kept its Nov 4 promise: Kimi-K3.1-Pro weights are on Hugging Face (2.8T MoE / 104B active, 1M context, multimodal). License is MIT-like with a MaaS $20M commercial gate and big-product attribution.
· MODEL Editors
Consent
Postman desktop auto-installs Codex/Claude plugins — without clear consent
HN ~1039 pts. Logs show auto-existing-upgrade installing agent plugins on open. Viral headlines overclaim 'share all API data' — MCP auth failed; no proven exfil. Postman says it will make opt-in clearer.
· MODEL Editors
FREE · EVERY WEEK
The top AI tools this week, tested.
Every week MODEL's editors pick the AI tools worth your time, try them hands-on, and tell you what's good, what's not, and who it's for. Free, in your inbox.
Free. Unsubscribe anytime. Our picks are editorial and can't be bought.