Alibaba announces Qwen2-VL AI model with advanced video analysis and reasoning capabilities

August 30, 2024: Alibaba Unveils Advanced Qwen2-VL AI Model - Alibaba Cloud announced Qwen2-VL, an AI model that excels in video analysis and multilingual conversations. Capable of understanding videos over 20 minutes long, it can summarize content, answer questions, and support live chat. Utilizing the Vision Transformer model with 600 million parameters and Multimodal Rotary Position Embedding, it outperforms many closed-source models. However, it cant extract audio and is trained up to June 2023. Available open-source, Qwen2-VL is a vital step toward comprehensive vision-audio AI models.