The apex-flash-1 model is post-trained on real-world vulnerabilities to assist specialists in identifying security flaws. Cantina Security developed the open weights model using a dataset of vulnerabilities its own researchers discovered and received payment for, aiming to provide a tool specifically tuned for security research.

A second abliterated variant of the model is also available to reduce refusals for users operating within authorized workflows. The release includes a technical guide detailing the construction and training process of the AI.