Tech giant Alibaba has announced the release of new open-source artificial intelligence (AI) models and text-to-video technology, intensifying its competitive stance in the generative AI space. The models, over 100 in total, belong to the company’s latest foundational language model series, Qwen 2.5, which was first introduced in May.
As both Chinese and U.S. tech companies race to develop AI tools, Alibaba’s approach stands out. Unlike rivals such as Baidu and OpenAI, which focus on proprietary, closed-source models, Alibaba has opted for a hybrid strategy. This involves investing in both proprietary and open-source initiatives, expanding its AI product offerings across sectors like automotive, gaming, and scientific research.
Alibaba’s new models vary in size from 0.5 to 72 billion parameters, making them versatile and capable of tasks such as mathematics and coding, while also supporting over 29 languages.
Alibaba also launched a new text-to-video AI model as part of its Tongyi Wanxiang image generation family, entering a space where competitors like ByteDance, the owner of TikTok, and OpenAI have already ventured.
