South Korea's Ministry of Science and ICT said on Jan. 23 it will accept applications until next month’s Feb. 12 to add one more elite team to the "indigenous AI foundation model" project.

As with the initial call, the ministry is seeking applications from elite teams centered on domestic AI companies and institutions. The project targets performance of at least 95 percent compared with the latest global AI models. Participating teams can present specific development strategies and methodologies.

The additional team must have technological competitiveness that can meaningfully compete with the three teams that advanced to the second-stage evaluation, including LG AI Research, SK Telecom and Upstage. The ministry explained that it must be capable of contributing to the growth and expansion of South Korea’s AI ecosystem, beyond simply taking on the challenge of developing an AI model.

The ministry plans to select an elite team that can compete with the existing three teams by proving its skills and capabilities across various indicators, including AI model development technology. It will also consider whether the team can contribute to the growth and expansion of the domestic AI ecosystem, as recognized by a majority of the expert evaluation panel.

If a majority of the evaluation panel judges that no team meets the criteria, the ministry may decide not to make an additional selection. Even if there is a single applicant, the selection evaluation will be conducted only for the applying consortium.

The ministry will support the selected additional team with 768 units of Nvidia’s advanced B200 GPUs, along with joint data purchasing and support for building and processing. It plans to provide support on the same level as existing elite teams, including granting the title of "K-AI company."

LG AI Research and Upstage, which previously advanced to the second evaluation stage, will receive support during the evaluation preparation process equivalent to 512 B200 GPUs and 512 H100 GPUs, respectively. SKT will receive GPU support starting from when it advances to the third-stage evaluation, as it is a government contractor for GPU leasing support.

The three elite teams will develop their indigenous AI models until June. The additional team will develop its model from February to July. This is intended to provide a development period equivalent to that of the other elite teams.

The second-stage evaluation is expected to take place around early August. The ministry will keep the broad framework of the existing stage evaluations, but plans to select global benchmarks targeting major global leaderboards and review measures such as strengthening the "independence" component in expert evaluation items.

A ministry official said it is an important time to create a bigger challenge through everyone’s efforts and move forward. The official stressed that the ministry will devote its full policy efforts to strengthening the competitiveness of South Korea’s AI ecosystem from a long-term perspective, not based on short-term results.