The upcoming open weight small Qwen models are likely going to lack training data related to offensive cyber capabilities.
retained text
With the release of Qwen 4 Max, they claim that training data related to offensive capabilities has been removed, is this largely a placebo/nothingburger or does it have the capability to genuinely lobotomize a model's capabilities in cyber?
document fetched 2026-11-18 00:46 · canonical https://reddit.com/r/LocalLLaMA/comments/1xykjuy/the_upcoming_open_weight_small_qwen_models_are