Cisco Launches Antares AI Models to Detect Vulnerable Code Locally Without Sending Data to the Cloud

Cisco launches Antares, a family of open-weight AI security models that detect vulnerable code locally without sending sensitive data to the cloud, outperforming larger models at a fraction of the cost.

• Cisco is launching Antares, a family of open-weight security small language models (SLMs) designed to pinpoint vulnerable code within codebases, with two models — Antares-350M and Antares-1B — now available on Hugging Face.
• Antares outperforms many larger closed- and open-weight models on vulnerability localization tasks at a fraction of the cost, and is compact enough to run locally, keeping sensitive source code off the cloud.
• Cisco also introduces a new 500-task Vulnerability Localization Benchmark to measure model performance on security-specific localization, filling a gap left by general coding benchmarks, as part of a broader push to make AI-assisted security accessible to under-resourced organizations.