Alibaba unveils new AI models and text-to-video tech

Alibaba, the Chinese tech giant, has intensified its push in the rapidly growing generative AI space by unveiling new open-source AI models and text-to-video technology. This move is part of the company's strategy to compete more aggressively in this booming field.

The newly released models, which come from Alibaba's Qwen 2.5 family of foundational large language models, include over 100 models of varying sizes, ranging from 0.5 to 72 billion parameters. These parameters help determine the model's overall performance and capability. Alibaba's models are designed to support a wide range of applications, spanning industries like automotive, gaming, and scientific research, and can handle tasks such as mathematics, coding, and support for over 29 languages.

While other companies like Baidu and OpenAI have primarily focused on closed-source AI developments, Alibaba has taken a more diversified approach, investing in both open-source and proprietary models. This strategy allows the company to expand its AI product offerings and compete on a global scale.

Additionally, Alibaba introduced a new text-to-video model under its Tongyi Wanxiang image generation family, positioning itself to compete with international players like OpenAI, which has also shown interest in the text-to-video space. This new model joins a wave of similar innovations from Chinese firms entering this market. In August, ByteDance, the parent company of TikTok, launched its own text-to-video app, Jimeng AI, for Chinese users on Apple's App Store.