Leaked 'DeepSeek-cyber' tease: open DeepSeek models near frontier on cybersecurity CVE benchmarks
teortaxesTex · x · 2026-10-30
teortaxesTex highlights a deleted post hinting at an internal "DeepSeek-cyber" model, arguing DeepSeek's cybersecurity knowledge has been strong since summer. Attached aikido/research benchmarks: DeepSeek V4.1 Flash found 84.4% of benchmark CVEs at pass@3, beating frontier models at 50x lower cost per task, with precision up to 78.9%; V4 Pro 0813 previously topped all models at 87.5% pass@3 but with only 65.6% precision.
More from Models
- Run Kimi K3 at home: tinygrad hits 40 tok/s interactive for about $30k — StephanSturges · 2026-10-30
- LLM redaction test at temp 0: some models leak hospital name every run, 24 runs cost $0.21 — MaziyarPanahi · 2026-10-30
- Daily AI Brief: Google kills Jules into Antigravity, OpenAI doubles Pro usage limits — testingcatalog · 2026-10-30
- Researcher: concept neurons alone may not be enough to induce an AI persona — Sauers_ · 2026-10-30
- Non-coder ditches Codex and ChatGPT for Kimi Work, says the experience is simply nicer — doctor_dadbod · 2026-10-30
- Anthropic's internal model found and exploited critical bugs in Chrome, Safari, Android and more — emmanuelvivier · 2026-10-30