today we're releasing apex-flash-1, our first open-weights model for security research post-trained on real vulnerabilities we found and got paid for.
we are also releasing the abliterated variant for researchers who want fewer refusals in their own authorized workflows.
how we built it: