SemiAnalysis: safety classifiers are underrated — Meta's Muse tops cyber benchmark at 35.2%
ThomasScialom · x · 2026-11-02
SemiAnalysis argues it's underrated how relevant safety classifiers are when picking a model for certain use cases.
The attached GDPx cybersecurity leaderboard (finding, exploiting and fixing proprietary zero-days in live multi-service environments) shows:
- Meta's Muse Wonder Full leads by a wide margin at 35.2%, with Muse Wonder second at 24.1%
- Gemini 4 Argon 18.5%, Gemini 4 Carbon 16.7%
- Open-weights DeepSeek 4.1 Flash matches at 16.7% for just $0.88/task
- GPT-6 Luna 11.1%, Claude Opus 5.5 only 9.3%; GPT-6 Astra and GPT 6.1 Sol score 0
Takeaway: model rankings on safety-heavy benchmarks diverge sharply from general leaderboards, and should factor into model selection.
More from Models
- OpenAI merges ChatGPT's chat and Work tabs; Codex now shares the same usage pool — TechCrunch AI · 2026-11-02
- Reddit users report Gemini Argon underperforming Gemini 3.1 Pro on instructions and audio — PreparationOdd430 · 2026-11-02
- Dev warning: avoid yearly prepaid LLM plans as labs quietly cut quotas — HankYeomans · 2026-11-02
- KVARK Sparkle 1.0 open-weights: Qwen3.8-27B derivative cuts thinking tokens by up to 67% — dzulajla98 · 2026-11-02
- 129 Cyber Incidents in 10 Days: Open Models Already Causing "Billions per Day" in Damages — Afinetheorem · 2026-11-02
- Rumor: Qwen 4 will be Alibaba's last open-weight model release — Chemical_Hawk_6307 · 2026-11-02