Cisco Launches Antares AI Models for On-Premises Software Vulnerability Detection
Cisco has introduced Antares, a new family of compact AI models designed to help organizations identify software vulnerabilities while keeping sensitive source code within their own IT environments.
Built as small language models (SLMs), Antares is designed to address one of cybersecurity’s most complex tasks—detecting hidden security flaws in software codebases—without requiring organizations to upload proprietary code to cloud-based AI services. The approach aims to reduce costs while meeting strict privacy, compliance, and data sovereignty requirements.
Cisco said the models can run entirely on-premises, allowing security teams to analyze code locally and maintain control over sensitive information.
“As regional organizations accelerate their digital capabilities, securing complex software without compromising data privacy is critical and urgent,” said Fady Younes, Managing Director for Cybersecurity at Cisco Middle East, Türkiye, Africa, Caucasus and Central Asia (METAC). “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.”
As part of the launch, Cisco is making two models—Antares-350M and Antares-1B—available as open-source resources for developers and cybersecurity professionals.
According to Cisco, Antares combines a compact design with advanced reasoning capabilities, enabling it to investigate software vulnerabilities in a manner similar to human analysts. Instead of relying solely on predefined rules, the models interpret vulnerability descriptions, search relevant sections of code, refine their analysis, and identify file paths most likely to contain security issues.
Cisco also claims benchmark testing shows the models outperform several larger AI models on key software security tasks while operating at significantly lower computational cost.
The company said the ability to deploy Antares locally makes it particularly suitable for government agencies, educational institutions, and organizations operating under strict data privacy or sovereignty regulations.
By releasing the models openly, Cisco aims to broaden access to AI-assisted software security, enabling smaller development and security teams to adopt advanced vulnerability detection capabilities without the infrastructure or costs typically associated with large language models.
With Antares, Cisco said it is seeking to support broader enterprise AI adoption by providing practical, privacy-focused tools that improve software security while reducing deployment complexity and operating costs.
