2 ms·Qwen3-Omni: first multimodal model with SoTA text, image, audio, and video perf2 points by walterbell 1y ago