All Four Korean Sovereign AI Models Win Global Recognition
2 min readSouth Korea just became the only country outside the United States and China where every active sovereign AI team holds a global benchmark distinction. The Ministry of Science and ICT announced this week that all four remaining teams in its sovereign foundation model competition have earned Notable Model status from Epoch AI, the research group whose registry tracks the world’s most significant AI systems.
The Toughest Rule in Sovereign AI
Korea’s program, launched in August 2025, set a bar few national AI efforts have dared: no foreign weights of any kind. Every architecture component, every training run and every model weight has to originate from domestic development. The rule has teeth. In January, Naver Cloud was eliminated after its submission used frozen encoder weights from Alibaba’s Qwen model for video and audio components, and NC AI was cut in the same round. Motif Technologies later joined through a supplementary evaluation, setting today’s field of four: LG AI Research, SK Telecom, Upstage and Motif.
What Happened
According to Tech Times, Epoch AI now lists models from all four teams as Notable Models, a recognition based on measures like state-of-the-art performance, scale and real-world significance. The milestone caps a frantic stretch for the program: two teams released open-source models in the 700 billion parameter class just days apart in early August, and a 200-person citizen panel selected by lottery spent August 8 to 11 evaluating all four systems ahead of the next elimination round.
Why It Matters
The stakes go beyond national pride. In December, the two winning teams become the technical suppliers for AI for All, a free national AI assistant intended for the country’s 51 million residents. Global recognition for all four finalists suggests the from-scratch requirement, widely criticized as too strict when Naver was cut, is producing models that stand on their own internationally.
For the many mid-sized nations weighing sovereign AI programs of their own, Korea’s experiment is turning into the reference case: strict rules, public accountability and now verifiable global results.
The elimination verdict comes next. Watch for one of the four to drop out before the December final.