I'm curious to see what other people's thoughts are.
But I'm very close to replacing claude code with qwen 3.8 27b model. A few more tests, but im about 99% sold on it.
When I saw the benchmarks comparing it to Opus 4.6. I didn't believe it. But I do now. This is soooooo good.
I'm calling it the "Diversity Ladder" for LLM vulnerability scanners. Maybe it's already a thing, I haven't seen it anywhere.
Welcome to my not Defcon talk: