Cisco Antares models have been introduced as a new family of small language AI models designed to detect software vulnerabilities on-premises. The company released these compact models openly to help organizations identify security flaws within their codebases without sending sensitive data to external cloud environments.
Scanning software code for vulnerabilities often presents financial and privacy challenges for modern enterprises. Traditional methods using general-purpose artificial intelligence typically require transferring proprietary source code to cloud servers, which can conflict with strict data sovereignty regulations.
The Capabilities of Cisco Antares models
Specifically, the Cisco Antares models family includes two open-weight versions, namely Antares-350M and Antares-1B, which are now available to the developer and security community. These compact models are designed to run locally on an organization’s own hardware, reducing the operational costs associated with larger cloud-based systems.
Privacy and Operational Efficiency
Consequently, running entirely on-premises allows public sector entities, academic institutions, and regulated industries to maintain complete control over their data, enhancing their overall cybersecurity posture. This local deployment model addresses compliance requirements while accelerating the speed of code analysis.
Notably, in benchmark testing, the compact models demonstrated the ability to perform critical security tasks at a lower cost compared to larger proprietary alternatives. The system operates by reading vulnerability descriptions, searching relevant codebases, and identifying the specific file paths most likely to contain security threats.
Regional Security Perspectives
“As regional organizations accelerate their digital capabilities, securing complex software without compromising data privacy is critical and urgent. With Antares, we are giving security teams the power of AI locally so they can pinpoint vulnerabilities faster while keeping sensitive source code firmly within their own secure environment.”
Fady Younes, Managing Director for Cybersecurity at Cisco METAC
Democratizing Software Security
The open release of these Cisco Antares models aims to provide smaller development teams with access to advanced security tools. Previously, deploying proprietary large language models required significant budgets and infrastructure that many smaller organizations could not support.
This release represents a shift toward establishing practical standards for local artificial intelligence deployment in the enterprise sector. Security teams can access the open-weight models to integrate them directly into their existing software development workflows, thereby improving overall digital resilience.
Furthermore, the availability of these tools helps bridge the gap between advanced threat detection and resource-constrained teams. By offering a local alternative, the initiative supports broader industry efforts to secure digital infrastructure against emerging threats.
