Alibaba has released Qwen2.5-Omni-7B, its first multimodal large language model capable of processing text, images, audio, and video simultaneously. This model can generate outputs in both text and natural-sounding speech in real time.

In benchmark tests, including OmniBench, Qwen2.5-Omni-7B outperformed several other models, such as Google’s Gemini 1.5 Pro. Alibaba states that the model can interpret multiple sensory inputs and respond effectively to complex tasks.

Recent Alibaba developments