Trillion Labs
Category: Foundation Models / LLMs
Korean foundation-model lab building specialized models for AI factories and industrial infrastructure (world models, energy/plant AI), plus open Korean Tri-series LLMs. Trillion Labs was founded in 2024. The company is led by Shin Jae-min (Jay Shin). Based in Seoul, South Korea. Team size: 11-50. Total funding raised: $4.2M. Latest round: Pre-Seed ($4.2M, Sep 2024). Key investors include Strong Ventures, Kakao Ventures, Goodwater Capital, Bass Investment, The Ventures, BAM Ventures.
- Founded
- 2024
- Headquarters
- Seoul, South Korea
- Team size
- 11-50
- Total funding
- $4.2M
Value proposition
Builds full-stack foundation models (LLMs, VLMs, world models) that understand and optimize physical AI infrastructure—AI factories, energy plants, industrial ops—via rBridge efficiency and Industrial World Models on NVIDIA stack.
Products and solutions
Tri-Series open Korean foundation models (0.5B–70B), Gravity MoE series (e.g. Gravity-16B-A3B), Industrial World Models (AI Factories / energy), gWorld mobile GUI world models (8B/32B), rBridge reasoning-proxy methodology, VLMs and diffusion LLM (dLLM)
Unique value
Full-stack foundation + world models that extend from efficient Korean/global LLM training (rBridge) to industrial intelligence for AI factories and physical infrastructure.
Target customer
AI factory / data center operators, energy and industrial infrastructure firms (e.g. GS Group collabs), government sovereign-AI programs, and enterprises needing Korean foundation models.
Industries served
AI Factories / Data Centers, Energy & Industrial Infrastructure, Generative AI / Foundation Models, Government & Public Sector, Enterprise Software (B2B)
Technology advantage
Efficient foundation-model training (rBridge) plus world models and full AI stack (train→ops) for industrial environments, integrated with NVIDIA Omniverse simulation and Nemotron reasoning.
How they differentiate
From-scratch foundation models (open Tri series) plus industrial world models that turn AI factories and energy plants into optimizable systems—not only Korean LLMs or generic chat apps.
Main competitors
Upstage, Naver Cloud (HyperCLOVA X), 42Maru
Key partnerships
NVIDIA (Inception; Omniverse + Nemotron Industrial World Models), AWS (Generative AI Accelerator), GS Group (energy infrastructure), Naver Cloud
Notable customers
GS Group (energy infrastructure collaboration), Undisclosed enterprise industrial POCs, Government AI / sovereign model initiatives
Major milestones
Founded 2024 (Seoul), Pre-Seed ~$4.2M (Sep 2024, Strong Ventures lead), AWS Generative AI Accelerator, Open-sourced Tri series including Tri-70B, gWorld paper accepted ICML 2026, Industrial World Models with NVIDIA Omniverse/Nemotron (Jun 2026), Released Gravity MoE models (2026)
Growth metrics
Open-sourced full Tri lineup (7B/21B/70B); Gravity MoE and gWorld (ICML 2026) on Hugging Face; Industrial World Models with NVIDIA announced Jun 2026; ~25–28 employees (TheVC/pension 2026).
Market positioning
Foundation-model lab for Industrial Intelligence / AI Factories (Korea base, global industrial GTM), with sovereign open Korean LLMs as first generation.
Geographic focus
South Korea, Japan, Southeast Asia
Patents and IP
Proprietary 'rBridge' algorithm and data synthesis pipelines.
About Shin Jae-min (Jay Shin)
Former Research Scientist at NAVER Cloud (HyperCLOVA X Team), focusing on pre-training and alignment of large language models. Previously an Applied Scientist Intern at Amazon Alexa AI. Ph.D. in Computer Science from KAIST.
Latest news about Trillion Labs
More Foundation Models / LLMs companies
Official website: https://trillionlabs.co
