🇻🇳Vietnam joins the AI model race: V-LLM launches, tops V-Bench ranking

On July 6, VinUni announced V-Bench - a benchmark set to evaluate the Vietnamese language capabilities of LLMs, built on over 40,000 questions. Initial results with 15 models:
- V-LLM v1 (Vingroup): 68.00 points - ranked first
- Gemini-2.5-Flash (Google): 58.80
- Qwen3-235B Thinking (Alibaba): 58.23
Vietnam officially has a name on the LLM map. Detailed score card
- Basic Politics: 99.7 - near perfect, highest in the table
- Logic: 32.9 - lowest in the top 7.
The model is nearly perfect on curriculum content, but reasoning is at the bottom. Along with that, Law 77.1 and Philosophy 76.8 - while Literature 34.3, Physics 40.1. Perhaps this is due to Vin's training focus.
Note that the opponents in the table are Gemini Flash version and open-source models, not GPT-5, Gemini Pro, or Claude.