Cantina Releases apex-flash-1, an Open-Weights Model for Vulnerability Research
Cantina Security and Yeta Labs have launched apex-flash-1, an open-weights AI model fine-tuned via reinforcement learning for vulnerability research. Built on GLM-5.3-Flash and released under an MIT license, the model successfully solved 40 out of 60 held-out bug tasks. This marks a significant step forward for open-source cybersecurity tools, though running it requires substantial GPU resources.
Source: MarkTechPost