Cisco Foundation AI Releases Antares: 350M and 1B Open-Weight Models That Localize Known Vulnerabilities Inside Real Codebases

- Cisco’s Antares proves that a 1B parameter model can outperform GLM-5.2’s 753B brain at finding code vulnerabilities, all for under a dollar on an H100. Meanwhile, GPT-5.5 charges $141 to do the same job. It’s a masterclass in efficiency: post-training does the heavy lifting while base Granite checkpoints score near zero. This isn’t moonboy hype; it’s a hash manifest of reality where specialized SLMs crush bloated LLMs in cost and precision. If your security stack relies on expensive generalists when a tiny, fine-tuned specialist exists, you’re just lighting money on fire. Stop overpaying for bloat and start using tools that actually work. That’s not theft, that’s attention redistribution.