South Korea eliminates Motif Technologies from sovereign AI competition, narrowing field to three

South Korea's Ministry of Science and ICT (MSIT) eliminated Motif Technologies from its sovereign AI competition, 'Dokpamo', after an evaluation round from August 8 to 11, leaving LG AI Research, SK Telecom, and Upstage as the final three contenders. Motif's 314B-parameter model failed to meet shifted criteria emphasizing industrial applicability and public usability, which included testing with 200 citizen evaluators. The remaining teams, equipped with approximately 768 NVIDIA B200 GPUs each, must provide at least half the inference capacity for a national free AI assistant for South Korea's 51 million citizens, with two final champions expected by late 2026 or early 2027.

Via thevc.kr South Korea eliminates Motif Technologies from sovereign AI competition, narrowing field to three The small startup's 314B-parameter model wasn't enough to survive evaluation criteria that shifted toward real-world usability over raw benchmarks. South Korea’s high-stakes sovereign AI competition just got smaller. Motif Technologies, a startup with roughly 30 employees, has been cut from the government’s “Dokpamo” contest after an evaluation round that ran from August 8 to 11, leaving only three teams standing: LG AI Research, SK Telecom, and Upstage. What happened to Motif Motif Technologies, a subsidiary of Moreh, entered the competition through a supplementary selection round in February 2026. It wasn’t part of the original lineup. The startup got its shot after two of the initial five competitors, Naver Cloud and NC AI, dropped out by January 2026, thinning the field and creating an opening. The company brought a credible offering to the table. Its Motif 3 model featured 314 billion parameters using a mixture-of-experts MoE architecture, along with proprietary mechanisms called GDLA and polynorm. In prior rounds, Motif had scored well on the Artificial Analysis Intelligence Index, a benchmark that measures raw model intelligence. But this latest evaluation round changed the rules of engagement. The Ministry of Science and ICT MSIT , which oversees the competition, shifted its criteria to emphasize industrial applicability, real-world capabilities, and public usability. That last piece included testing with 200 citizen evaluators, regular people assessing how well these models actually work in practice. The three survivors and their models LG AI Research leads the pack in raw scale with its K-EXAONE 2.0 model at 750 billion parameters. SK Telecom’s A.X K2 isn’t far behind at 688 billion parameters. Upstage rounds out the trio with Solar Open 2 at 250 billion parameters. Each team has been equipped with approximately 768 NVIDIA B200 GPUs. The remaining contenders are required to provide at least half the inference capacity for a national free AI assistant, a public utility available to all of South Korea’s 51 million citizens. Two final champions are expected to be selected by late 2026 or early 2027. The bigger picture: sovereign AI as national strategy The competition originally launched with five teams in August 2025. The early dropout of Naver Cloud and NC AI hinted at how demanding the process would be. South Korea has been building a legislative framework around this ambition, including the AI Basic Act, which is designed to support domestic AI development while establishing governance guardrails. Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy https://cryptobriefing.com/editorial-policy/ .