Cisco engineer replicates free Antares security model at ~80% of official score
shashib · x · 2026-10-24
Cisco Principal Engineer Julio Gomez ran the free Antares-1B security model through 500 cases from Cisco's own benchmark in 9 hours on a MacBook, scoring 0.168 vs Cisco's published 0.209 — about 80% by the author's arithmetic. Antares reads new security advisories and ranks which files in a codebase to open first, replacing 2 hours of keyword searching per 300-file project. Gomez posted results on GitHub for anyone to rerun; cost was $0.71 vs $141 for GPT-5.5 on the same 500 cases, and GPT-5.5 only overtakes Antares on projects above 10 MB. No independent lab has yet replicated the results.
More from Models
- Scale CEO teases new models as users say Muse Wonder Full is underrated — alexandr_wang · 2026-10-24
- Canva Craft trained on millions of templates and 25,000 designer-evaluated image pairs — jacob_top · 2026-10-24
- $34.72 of Kimi K3.1 Pro tokens builds a playable Spider-Man Manhattan in 8 hours — sven_ai · 2026-10-24
- Kimi K3.1 vs Opus 5.5: 80% Cheaper and 4x Faster on Same Task — sven_ai · 2026-10-24
- Claim: Open-source Kimi K3.1 Pro beats OpenAI's GPT-6-Astra, unverified — sven_ai · 2026-10-24
- Moonshot dev explains why Kimi K2 went open-source — and why ads stopped — teortaxesTex · 2026-10-24