Coding via Video: Alibaba Unveils Qwen3.5-Omni — The AI That "Sees" Your Screen

The Qwen team (Alibaba Cloud) has launched Qwen3.5-Omni, a next-gen multimodal model that processes text, images, audio, and video in real-time. Its most standout feature, "Audio-Visual Vibe Coding," allows the AI to watch a screen recording with audio instructions and generate functional code without any text prompts.

Qwen3.5-Omni has surpassed Gemini 3.1 Pro in audio understanding and matches it in video processing. Supporting 113 languages and featuring a 256k token context window, it can analyze over 10 hours of audio in a single request. Its new ARIA technique ensures flawless speech synthesis, making it a formidable competitor for ElevenLabs and GPT-Audio.

This material was prepared by the "Amul Info" tech desk based on an analysis of the Hybrid-Attention MoE architecture by Alibaba Cloud.