To spell it out: if you are not in the CVP or Daybreak program, the best cybersecurity model in the world is Chinese. It is not in 6th place behind Mythos, Fable, Astra, Opus, and Sol. It is in 1st.
And even if you are in these programs, you still get plenty of downgrades from the classifier, and rejections from the model
There are big implications to this. Open research being biased towards GLM’s capability frontier. More production traces going to ZAI. More tooling built around open models in general. Etc