Cisco Foundation AI Releases Antares: 350M and 1B Open-Weight Models That Localize Known Vulnerabilities Inside Real Codebases

Cisco Foundation AI released Antares, a family of open-weight security small language models for vulnerability localization.

The company also open-sourced the 500-task VLoc Bench benchmark and two Apache 2.0 models: Antares-350M and Antares-1B.

Evaluated on real repositories tied to public security advisories, the 1B model achieved a 0.209 File F1 score.

Cisco says the specialized models can localize vulnerable files competitively with much larger general-purpose models and some static analysis tools, potentially lowering triage costs.