Trillion Labs, a leading AI foundation model startup, was featured as a key global partner during the keynote speech at NVIDIA GTC 2026, the world’s premier AI conference. This recognition positions Trillion Labs as a primary hub for ‘Sovereign AI’ in South Korea.

Jensen Huang Identifies Trillion Labs as a Strategic Partner for South Korea’s AI Sovereignty

During his keynote at GTC 2026 held on March 16 (local time) in San Jose, California, NVIDIA CEO Jensen Huang presented the current progress of building regional AI ecosystems based on NVIDIA Nemotron. On the global map showcasing strategic regional hubs, Trillion Labs was prominently featured as a major partner for South Korea.

Nemotron is NVIDIA’s family of open models specifically designed for advanced reasoning, coding, and agentic tasks. It serves as a foundational platform for developing a wide range of generative AI systems, including Large Language Models (LLMs), Vision Language Models (VLMs), video models, and speech AI.

This recognition provides global validation that Trillion Labs is operating as a critical base for safeguarding national AI sovereignty. By leveraging its proprietary technical expertise within NVIDIA’s latest open-model ecosystem, Trillion Labs continues to lead the advancement of independent, high-performance AI solutions for the region.

Recognized as an official partner by NVIDIA validates that Trillion Labs' proprietary AI architecture and technical capabilities have met the highest global standards. This partnership signifies more than just technical cooperation; it positions Trillion Labs as a critical hub for demonstrating the technological independence of "K-AI" within NVIDIA’s cutting-edge global ecosystem.

‘The Frontier Lab’: Redefining AI Paradigms within 18 Months of Founding

Launched in August 2024 as the only startup in Korea with full-stack, from-scratch LLM development capabilities, Trillion Labs has achieved unprecedented technical milestones in a remarkably short period.

• Proven Global Performance & Efficiency: Starting with 'Trillion-7B', a high-efficiency multilingual LLM, the company became the first Korean startup to complete a 70B-parameter flagship model with full checkpoint transparency, setting new global standards for research reliability. Recently, its latest model, 'Tri-21B Think', ranked among the top 30 globally on the Artificial Analysis leaderboard, proving that its proprietary architecture can outperform significantly larger models without the need for massive corporate infrastructure.
  
  

• Next-Gen Reasoning with ‘Backtracking’: 'Tri-21B Think' implements a unique Backtracking structure that allows the AI to review and self-correct its reasoning process before generating a final answer. Powered by 'Test-Time Scaling' technology, the model’s problem-solving capabilities scale exponentially with compute time, outperforming models with over 100B parameters even when running on a single GPU.
  
  

• Leading Technological Evolution: By unveiling 'Trida', a diffusion-based LLM that replaces traditional auto-regressive methods, and 'gWorld-32B', the world’s first mobile World Model, Trillion Labs is establishing a continuous evolutionary path from Diffusion LLMs and VLMs to World Models that predict action outcomes.
  
  

• National Recognition & Trust: Based on these achievements, the lab is a key participant in the Ministry of Science and ICT’s 'AI Foundation Model Project' for healthcare and science. Its domain-specific MoE (Mixture of Experts) technology has been designated as a National Strategic Technology, providing a foundation for trustworthy AI in public, defense, and medical sectors.