The company is open-sourcing two small-language models designed just to hunt for software bugs.

The company is open-sourcing two small-language models designed just to hunt for software bugs.

Cisco Foundation AI has announced Antares, a small language model (SLM) designed for low-cost vulnerability hunting.

Cisco Foundation AI released Antares, open-weight 350M and 1B models that localize known vulnerabilities across real code repositories.

The new models perform competitively with GPT-5.5 on a security benchmark but at a much smaller fraction of the cost.

Cisco releases small, specialized language models optimized for CVE hunting. They work 4.5× faster than GPT-5.5 and 141× cheaper.