Zalo’s LLM (large language model) is in competition with other LLMs in the market. Zalo, a Vietnamese company, has developed their own LLM model in the Vietnamese language. The aim of this model is to generate AI apps that can generate compatible documents and better serve Vietnamese users.
At the recent AI Summit 2023, Zalo showcased their LLM, which is still in the development phase. The model demonstrated the ability to communicate and answer user questions in different fields in the Vietnamese language. To test the model’s capabilities, it was integrated into Kiki for a knowledge competition via the Kahoot game. Other previously released LLM models such as GPT-3.5, GPT-4, Llama22, and QWen, a recently released Vietnamese LLM, also participated in the competition.
The competition consisted of 20 questions from various fields, and the performance of the LLMs in answering these questions was displayed on a screen for transparency. Zalo’s LLM performed well in answering general knowledge questions, along with Kiki and GPT-4. However, when faced with questions about maximum vehicle speed in urban areas and the animal designation of 2024, all LLM models struggled to answer correctly.
Overall, Zalo’s LLM proved to be on par with other LLM models from Meta and Alibaba, showing promise for the development of a good LLM model for the Vietnamese language. While the LLMs were unable to defeat real players in the competition, this demonstrates the ongoing advancements in LLM technology and the potential for further development in the future.