Benchmarking open-weight models for security research

GLM 5.1 is surprisingly capable. Anecdotally, I couldn't notice a difference until ~120K tokens.

Qwen 3.6 35B A3B also exceeded my expectations. It's surprisingly performant, even though the previous generation wasn't even able to use the testing harness.

(Tbd on Kimi K2.6; the eval is still running.)