Cisco introduces Antares AI models, compact in size and locally used to detect software vulnerabilities

Cisco has introduced Antares, a family of small language models designed specifically to detect security flaws hidden in software source code, locally, without the need to externalize the data.

Antares is a new family of artificial intelligence models whose name is inspired by a red supergiant star, the brightest in the constellation of Scorpio.

Cisco has made it openly available to the developer and security community with two variants: Antares-350M and Antares-1B. A third, Antares-3B, will be available soon.

Both are small in size, which makes it easier for them to be run locally within the organization’s secure environment and not in the cloud, since the latter can put confidential and proprietary source code at risk, as reported in a press release.

They are designed to strengthen the security of organizations by reading vulnerability descriptions, searching for relevant code, changing addresses if a path is not useful, and narrowing down file paths most likely to contain a threat.

Cisco claims that despite their size, the Antares models outperform many larger, more expensive AI models in critical security tasks, performing at a fraction of the cost.