On November 25, 2025, Singapore's National Artificial Intelligence Initiative (AISG) made a significant announcement. In its most recent large - model project focusing on Southeast Asian languages, the initiative decided to ditch Meta's model and embrace Alibaba's Tongyi Qianwen (Qwen) open - source architecture. This move led to the launch of the "Qwen - SEA - LION - v4" model. Remarkably, this model swiftly ascended to the pinnacle of an open - source leaderboard that evaluates capabilities in Southeast Asian languages.

In the past, open - source models like those in Meta's Llama series struggled when dealing with regional languages such as Indonesian, Thai, and Malay. Their subpar performance significantly hampered the development efficiency and overall performance of localized AI applications. The decision by AISG to switch models not only resolves this long - standing issue but also serves as a crucial indicator of the growing global influence of Chinese open - source AI models.