Overview
- LG AI Research led the first-stage review with 90.2 points and topped benchmark, expert and user evaluations, according to the Science and ICT Ministry.
- The assessment weighed benchmark testing at 40 points, expert panel reviews at 35 and user feedback at 25, with auditors examining technical reports and training logs.
- Naver Cloud was dropped over originality concerns linked to reliance on overseas model elements and pre-trained weights, and NC AI also failed to advance.
- Officials plan to add one more consortium to the next phase, with selected teams receiving high-performance compute, data access and a formal K-AI designation.
- Naver, Kakao and NC AI have indicated they will not seek reentry in the additional round, prompting questions about the depth of competition going forward.