Qwen3 VL support merged into llama.cpp

Benchmarks look pretty good, even better than some of the text only models, make sure to take them with a grain of salt tho

Benchmarks