Bottom line: Cisco has released “Antares,” a family of open small language models for source code vulnerability analysis that the vendor says cost less than one US dollar per analysis.
Cisco is introducing “Antares,” a family of small language models designed to locate vulnerabilities in source code. The models are available as an open-weight version on Hugging Face and, according to the vendor, can be used for less than one US dollar per analysis.
Cisco has released a series of small language models (SLMs) under the name “Antares” that specialize in identifying security vulnerabilities in source code. The models are available as an open-weight version on the Hugging Face platform, allowing companies to run them themselves and integrate them into their own review processes.
For CISOs, the cost aspect is particularly relevant: Cisco states that an analysis using the Antares models costs less than one US dollar. This positions the offering as a low-cost alternative to larger, more resource-intensive language models used for comparable tasks in source code analysis. Especially when scanning large codebases or performing frequently repeated automated checks as part of CI/CD pipelines, the cost difference can become significant at scale.
The fact that the models are available as open weights allows for local operation without connecting to third-party cloud services. This can be important for organizations with high requirements for data protection and source code confidentiality, since sensitive code does not need to be transmitted to external API endpoints. At the same time, it remains unclear how the detection accuracy of the SLMs compares to established static application security testing (SAST) tools or larger foundation models, as the original source does not provide any benchmark data on this.
Before productive deployment, security officers are advised to conduct their own evaluation using known vulnerability classes and existing codebases in order to realistically assess false-positive and false-negative rates before integrating Antares into existing application security workflows.