[Alibaba Qwen 3.5-Omni Multimodal Model Launched] Alibaba Qwen has announced the launch of its multimodal model Qwen 3.5-Omni. The Qwen 3.5-Omni series includes Instruct versions in three sizes: Plus, Flash, and Light, supporting a 256k long context. The model supports over 10 hours of audio input and more than 400 seconds of 720P (1FPS) audio-visual input. It has undergone native multimodal pretraining on massive amounts of text, visual data, and over 100 million hours of audio-visual data, showcasing exceptional multimodal perception and generation capabilities. Compared to Qwen 3-Omni, Qwen 3.5-Omni has significantly enhanced multilingual capabilities, supporting speech recognition in 113 languages and dialects and speech generation in 36 languages and dialects.