Trillion Labs
3 known investors
Trillion Labs develops specialized foundation models designed to build and operate the physical infrastructure — energy, chips, and facilities — of AI factories that train AI. The models are aimed at automating and improving the operation of AI data center infrastructure.
Also known as TrillionLabs
Investors · 3
Company profile
researched Aug 2026Trillion Labs is a Seoul-based foundation-model company building specialized models for the physical layer that AI systems run on — energy, chips and data-center infrastructure, which it frames as "AI factories." Its stated mission is recursively self-improving AI infrastructure, in which models operate and improve the facilities that train subsequent generations of models.
The company's research portfolio spans several model families. The Tri series (roughly 0.5B to 70B parameters) comprises open Korean foundation models trained from scratch and represents its first generation. Gravity is a newer mixture-of-experts series, including Gravity Nano (16B-A3B), aimed at STEM reasoning and agentic tasks. rBridge is a methodology for predicting frontier reasoning performance from a small proxy model. Additional work covers vision-language models, a diffusion LLM, gWorld mobile GUI world models (8B and 32B), and Industrial World Models targeted at AI factories and energy plants.
Applied products include AI agents and models for energy plants, with industrial deployments developed on the NVIDIA stack using Omniverse simulation and Nemotron reasoning.
Business model
Trillion Labs develops foundation models in-house and releases part of its model lineup openly (the Tri series, Gravity MoE models and gWorld models are distributed publicly, including via Hugging Face), while positioning industrial world models and AI-factory/energy applications toward enterprise, infrastructure and public-sector buyers.
Traction
The full Tri lineup (7B/21B/70B) has been open-sourced, and Gravity MoE and gWorld models are published on Hugging Face. Tri 21B Think was reported in the global top 30 on the Artificial Analysis leaderboard. Reported headcount is approximately 25-28 (team size band 11-50), with named collaborations including NVIDIA, AWS, Naver Cloud and GS Group, plus undisclosed enterprise industrial proofs of concept and government AI initiatives.
Latest developments
The company announced its Gravity model series, including Gravity Nano (16B-A3B), a from-scratch mixture-of-experts line for STEM and agentic reasoning. In June 2026 it announced Industrial World Models built with NVIDIA Omniverse and Nemotron. Its gWorld mobile GUI world model paper was accepted to ICML 2026, and Tri 21B Think — a 21-billion-parameter model using reinforcement learning and test-time scaling — was reported to have entered the global top 30 on the Artificial Analysis leaderboard. The company is hiring across AI research, infrastructure engineering, business development and finance operations.
▸Full profile — market position, technology, go-to-market, geography, history
Market position
Positioned as a foundation-model lab for industrial intelligence and AI factories, based in Korea with a global industrial go-to-market, using sovereign open Korean LLMs as its first generation. Named competitors include Upstage, Naver Cloud (HyperCLOVA X) and 42Maru.
Trillion Labs trains foundation models from scratch and releases them openly (the Tri series), and extends beyond language modeling into industrial world models intended to make AI factories and energy plants optimizable systems rather than serving only Korean-language LLM or general chat use cases. It also cites proprietary methodology in rBridge and its data-synthesis pipelines.
Technology
Full-stack foundation models trained from scratch, including LLMs, vision-language models, a diffusion LLM, mixture-of-experts architectures (Gravity, e.g. 16B-A3B) and world models (gWorld GUI world models at 8B and 32B; Industrial World Models). The rBridge method predicts frontier reasoning capability from a small proxy model to improve training efficiency. Industrial deployments are integrated with NVIDIA Omniverse simulation and Nemotron reasoning. The company also cites proprietary data synthesis pipelines.
Go-to-market
The company pairs open model releases and published research (Hugging Face distribution, ICML 2026 acceptance, Artificial Analysis leaderboard placement) with partnerships and accelerator programs — NVIDIA Inception, the AWS Generative AI Accelerator, Naver Cloud, and an energy-infrastructure collaboration with GS Group — alongside enterprise industrial proofs of concept and government sovereign-AI initiatives.
AI factory and data center operators, energy and industrial infrastructure companies, government sovereign-AI programs, and enterprises requiring Korean foundation models.
Geography
Headquartered in Seoul, South Korea, with stated market focus on South Korea, Japan and Southeast Asia and a global industrial go-to-market orientation.
History
The company was founded in 2024 in Seoul and first built the Tri series of open Korean foundation models spanning roughly 0.5B to 70B parameters. It was selected for the AWS Generative AI Accelerator and joined NVIDIA Inception. Subsequent work extended from language models to reasoning-efficiency methodology (rBridge), gWorld mobile GUI world models accepted at ICML 2026, Gravity MoE models released in 2026, and Industrial World Models announced with NVIDIA in June 2026.
Compiled by commissioned research from 4 cited public sources — announcements, filings, and press listed under research sources below.
Key figures
latest reportedCompany-reported or press-reported figures, each dated to when it was claimed — not independently audited.
Timeline · 7
launches, deals, and filingsAnnounced Industrial World Models for AI factories and energy infrastructure developed on the NVIDIA stack, integrating Omniverse simulation and Nemotron reasoning; the company is also an NVIDIA Inception member.
Research on gWorld mobile GUI world models (8B/32B) accepted to ICML 2026 and published on Hugging Face.
A 21-billion-parameter model using reinforcement learning and test-time scaling placed in the global top 30 of the Artificial Analysis leaderboard.
Collaboration applying the company's models to energy infrastructure.
Trillion Labs announced its Gravity model line, a mixture-of-experts series trained from scratch and positioned for STEM reasoning and agentic tasks, including Gravity Nano (16B-A3B).
Released the Tri family of open Korean foundation models trained from scratch, spanning roughly 0.5B to 70B parameters, including Tri-70B.
Dated company events from announcements, filings, and press; legal rows summarize public dockets and regulator releases.
In the news
▸Research sources · 4
primary sources listed
- Trillion Labstrillionlabs.co · web
4 public sources were cited for this profile; the first-party ones are listed here.
Frequently asked questions
- What does Trillion Labs do?
- Seoul-based lab building foundation and world models that operate AI-factory infrastructure — energy, chips and facilities.
- Who are Trillion Labs's investors?
- Trillion Labs's investors include BAM Ventures, Kakao Ventures, TheVentures.