Finding vulnerabilities in code is one of the hardest problems in software. As an industry, we need multiple ways to identify and fix them. That's why we're introducing the #Antares family of SLMs. Today we're making Antares-350M and Antares-1B available as open-weight models on @huggingface. People have asked whether these models compete with frontier models. They don't. Antares solves a different problem. It is purpose-built for security, and within security, specifically optimized for finding vulnerabilities in code. It gives customers a specialized model they can run at a lower cost, more control, and in environments where their code never has to leave their infrastructure. Frontier models are extraordinary and will continue to push the boundaries of what's possible. We use them extensively and will continue to partner deeply with frontier labs. Huge congratulations to the research team behind this work. @djsampath, @aminkarbasi & team, fantastic job on taking yet another step at keeping the world safe. Security is a team sport. The more ways we can help developers find vulnerabilities before attackers do, the safer the entire ecosystem becomes.

This is going to be very fun to share with everyone tomorrow. And important for the world. Go @djsampath and @aminkarbasi !