SOTA 1 πΌοΈ Qwen2.5-VL: Next-Gen Vision-Language Model with Dynamic Resolution & Long Video Understanding Sep 8, 2025