GreenNode - the "flagship" unit specializing in AI Cloud of VNG introduced the first open-source Vietnamese LLM large language inference model on the NVIDIA NIM platform; while Zalo AI's representative brought practical lessons to develop a "sovereign" LLM associated with the characteristics of the Vietnamese language.
At NVIDIA AI Day Ho Chi Minh City, Mr. Le Hong Minh, Founder and Chairman of VNG, shared initiatives that contribute to building and perfecting a sovereign AI platform for Vietnam. GreenNode - the "flagship" unit specializing in AI Cloud of VNG introduced the first open-source Vietnamese LLM large language inference model on the NVIDIA NIM platform; while the representative of Zalo AI brought practical lessons to develop a "sovereign" LLM associated with the characteristics of the Vietnamese language.
NVIDIA AI Day Ho Chi Minh City is a series of large-scale events organized by NVIDIA in many cities around the world, allowing attendees to experience the latest AI technologies, connect and share strategies to bring AI from theory to practice. After Ho Chi Minh City, the event will be held in Tokyo, Sydney and Seoul.
Accompanying Vietnam in building a sovereign AI platform
At the special discussion session with the topic "Building Sovereign AI for Vietnam: Vision, Capacity and Opportunities", Mr. Vo Xuan Hoai - Deputy Director of the National Innovation Center (NIC), Mr. Le Hong Minh - Chairman of VNG and Mr. Pham Minh Tuan, CEO of FPT Software shared and analyzed strategies that can help Vietnam strengthen its technology infrastructure, develop an open AI ecosystem, pioneer innovation and build a sovereign, sustainable digital economy.
Specifically, building high-performance cloud computing infrastructure, developing standardized open datasets, building an AI startup ecosystem, cultivating AI talent, and perfecting policies are top priorities to promote Vietnam's sovereign AI.
“The Government is focusing on building a Vietnamese data set to serve AI research and development. In addition, developing an AI talent pool is the most important priority, which needs to be implemented immediately and effectively with the goal of having more than 50,000 engineers in the coming years; at the same time, the Government needs to issue comprehensive support policies for businesses, startups and the human resources community, thereby forming an AI startup ecosystem within the next 5 years, ” said Mr. Vo Xuan Hoai.
Currently, promoting the development of artificial intelligence (AI) is identified as one of the key tasks of the Party and Government, demonstrated through the National Strategy on AI research, development and application to 2030, with the goal of making Vietnam a leading center for innovation and AI application deployment in the ASEAN region.
VNG Chairman Le Hong Minh shared: “ Sovereign AI must be able to build real AI applications and business models with distinct values, in the context that Vietnam's investment capital is still very small compared to the world. Besides, we still lack a team of AI research experts and the accumulation of core technology platforms. To solve these two problems, Vietnamese technology enterprises need to focus on developing and perfecting products and specific business opportunities. Users cannot be "forced" to use AI, but AI must be naturally integrated into the experience.”
For AI, deployment speed plays a key role. VNG successfully commercialized AI Cloud within 6 months, 20% of Zalo users are currently using AI features after 2 years of testing. In addition, VNG is always looking for opportunities to cooperate with the Government, universities, researchers and startups to develop applications of high practical value.

“AI today is like the Internet in 1995-1996: there are many doubts and uncertainties. Developers need to stay excited about the possibilities that technology brings, and not take AI for granted. Learn, boldly explore, try and make mistakes, and stick to the revenue momentum from the beginning. Unlike Silicon Valley, where startups can raise capital even without a product, in Vietnam, you have to both explore and find a way to survive,” Mr. Le Hong Minh emphasized.
The first Vietnamese Large Language Model (LLM) on NVIDIA NIM
During the event, GreenNode introduced GreenMind-Medium-14B-R1 (GreenMind). This is the first open-source Vietnamese LLM inference model integrated on the NVIDIA NIM platform and optimized to run on a single NVIDIA H100 Tensor Core GPU.
Thanks to its self-improving data loop mechanism, combined with human participation throughout the training process, GreenMind is suitable for many highly practical applications such as: enterprise AI assistant, smart chatbot, advanced document retrieval and information search system or complex reasoning tasks in Vietnamese.
Sharing about GreenNode's vision, Mr. Vo Trong Thu - Head of AI Lab, GreenNode said: "GreenNode focuses on mastering core technology, while developing GreenNode AI Platform as an open platform, making AI more accessible and closer to the community. The launch of GreenMind on NVIDIA NIM marks an important step forward, helping Vietnam to be more proactive in developing native AI models, while creating a platform for domestic businesses and startups to apply AI flexibly, cost-effectively and in accordance with practical needs."

Developed and trained by a team of Vietnamese AI engineers, GreenMind is an accessible domestic model with an open source orientation. This model not only helps optimize costs and comply with regulations, but also flexibly integrates into existing infrastructure to support Vietnamese businesses in exploiting advanced technology, reducing dependence on foreign resources.
“ The new generation of AI models does not just stop at answering machine questions, but begins to 'think': understand context, distinguish concepts and give more accurate answers. To do that, we must build reasoning data (multi-step reasoning) in Vietnamese, trained on the knowledge base of Vietnamese people - from professional knowledge to culture, language, idioms. Only then can we create AI models that are truly dedicated to Vietnamese people, accurately reflecting Vietnamese thinking and values, ” Mr. Vo Trong Thu explained.
Invest heavily in core AI capabilities
Besides the orientation of developing open-source LLM, VNG is still investing heavily in LLM training from scratch, aiming at the goal of AI mastered by Vietnamese people to be able to integrate into global trends.
Dr. Chau Thanh Duc, Head of Research and Development at Zalo AI, emphasized that developing a large, "sovereign" language model associated with the characteristics of Vietnamese requires quality data, the ability to deploy on a large scale, and training methods suitable for limited resource conditions.

These are the practical experiences that Zalo has drawn from its journey to build AI products specifically for Vietnamese people, realizing the vision of "AI-First". Currently, Zalo AI is operating specialized research and development labs, serving the research and application of many advanced AI technologies.
With the “from-scratch” technique - training from zero, the Vietnamese LLM model developed by Zalo's engineering team has achieved 1.5 times the capacity of OpenAI's GPT-3.5 - according to VLMU ( Vietnamese Multitask Language Understanding Benchmark Suite for Large Language Models ). At the end of 2024, this model ranked second only to Meta's LLaMA-3-70B in terms of Vietnamese processing capacity on VMLU's ranking.
This success of Zalo has made Vietnam one of the few Southeast Asian countries to own a domestic LLM model, marking a strategic milestone in the context of increasingly fierce global technology competition.
Most recently, Zalo has officially launched the AI Citizen Assistant, helping 78.3 million regular Zalo users look up information on administrative procedures and current legal regulations conveniently, quickly and accurately through AI. Integrated right on the Zalo platform, this application contributes to realizing the goal of every citizen having an AI "assistant".
In addition, AI features on the Zalo platform such as text-to-speech, message translation, voice typing... have attracted more than 20% of Zalo users to experience regularly every month. " Zalo's big goal is that in 5 years, 100% of Vietnamese people will use the messaging platform and smart AI assistant developed by Zalo ", Mr. Le Hong Minh shared.
The Sizing the Prize (PwC) report shows that AI could contribute $15.7 trillion to global GDP by 2030. According to the WIN World AI Index 2025, Vietnam is currently ranked 6th out of 40 countries, scoring 59.2 points on a 100-point scale, in terms of AI readiness.
AI is identified as a key field, forecast to contribute about 80 billion USD - equivalent to 12% of GDP to Vietnam by 2030 if widely applied (according to Google). It is expected that by the end of 2025, Vietnam will announce an updated National AI Strategy along with the first AI Law.
Source: https://nhandan.vn/vng-group-hien-ke-giai-phap-xay-dung-ai-co-chu-quyen-cho-viet-nam-tai-nvidia-ai-day-ho-chi-minh-city-post909977.html
Comment (0)