OpenAI teases cyber model scoring 79.1% on PatchBench, beating GPT-6
OpenAI teased an upcoming cyber model scoring 79.1% on internal PatchBench, beating GPT-6, though researchers mocked the misleading chart used in the announcement.
2026-11-13 ~ 2026-11-13 · 2 related posts
- Episode 1: OpenAI's autonomous pipeline validates and patches 948 Linux kernel LPE bugs(2026-11-05, 28 posts)
- Episode 2: AI Vulnerability Disclosure Now Rivals Traditional Software, Says OpenAI Researcher(2026-11-06, 2 posts)
- Episode 3: OpenAI Launches GPT-6 Cyber, a Restricted-Access Cybersecurity Model(2026-11-06, 33 posts)
- Episode 4: Leaked GPT-6 Cyber Security Model Scores 48.5% on ExploitGym(2026-11-06, 3 posts)
- Episode 5: OpenAI's unreleased cyber model scores 77% on ExploitGym(2026-11-13, 7 posts)
- Episode 6: OpenAI teases cyber model scoring 79.1% on PatchBench, beating GPT-6(2026-11-13, 2 posts)
- OpenAI Teases Upcoming Cyber Model Scoring 79.1% on Internal PatchBench, Beating GPT-6 Cyber — JiaweiLiu_ · 2026-11-13
- OpenAI's upcoming Cyber model hits 79.1% on PatchBench, mocked for "cyber chart crimes" — moyix · 2026-11-13