SyntheticGestalt
  • Top
  • Technology
  • Services
  • Culture
  • Company
  • Careers
  • Contact
  • News
  • Twitter
  • Facebook
  • Linkedin
  • Tokyo Office

    11F H1O Hamamatsucho, 1-2-14 Shibadaimon, Minato City, Tokyo, 105-0012

Contact
Privacy Policy

© SyntheticGestalt KK

Previous
Next

SyntheticGestalt presented development achievements of the world's largest molecule-specialized foundation AI model "SG4D10B" at CEATEC 2025!

2025-10-23
template_master_creation.png

SyntheticGestalt (Shinjuku, Tokyo; CEO Koki Shimada; hereinafter "our company") is pleased to announce that it presented the development achievements of its world's largest molecule-specialized foundation AI model "SG4D10B *1" at CEATEC 2025 *2, held from October 14-17 this year.


CEATEC is the Asia's largest comprehensive technology exhibition, attracting 100,000 visitors and over 800 exhibitors from more than 20 countries worldwide. It brings together cutting-edge technologies across diverse fields including AI, IoT, DX, robotics, and mobility, and has garnered significant attention as a venue for cross-industry collaboration and business matching.


Our CTO, Kotaro Kamiya, presented on the theme "Development Achievements of Foundation Model Development Companies in GENIAC *3," sharing the results achieved through our participation in GENIAC *4 cycle 2:

  • Developed "SG4D10B," the world's largest molecule-specialized foundation AI model trained on 10 billion compound data points
  • Achieved the world #1 rankings in three critical drug discovery benchmark indicators (toxicity, permeability, and stability)
  • Released the compact model "SG4D100M," pre-trained on 100 million compound data points, for free on Google Cloud Marketplace and AWS Marketplace
  • Published 7,770 fragment compound activity data points obtained through wet-lab experiments in a public repository on GitHub

In recent years, AI-driven drug discovery has gained significant attention, with many companies developing AI models. However, due to limited data volumes and the complexity of molecular information, molecule-oriented AI has not achieved sufficient accuracy. SG4D10B solves this challenge through the world's largest-scale training dataset and proprietary architecture. It enables highly accurate compound activity prediction, contributing to shortened development timelines and cost reduction.


Additionally, Daisuke Morohashi, Head of AI Business, presented on the theme "Japanese Companies' Deployment Potential and Overseas Opportunities *5." Together with speakers from the Ministry of Economy, Trade and Industry, Japan International Cooperation Agency, and Preferred Networks, he participated in discussions on topics including "opportunities and current status for domestic generative AI companies expanding overseas," "hurdles when expanding overseas and how to overcome them," and "overseas expansion prospects from the perspective of domestic generative AI companies," incorporating real examples from our company.


About SyntheticGestalt

SyntheticGestalt develops AI technology specialized in molecular information. In 2024 and 2025, we unveiled the world's largest molecule-specific foundation model at NVIDIA GTC (one of the world's largest AI conferences). Our molecular foundation model resolves accuracy challenges faced by conventional molecular AI. We aim to apply this technology across various fields including pharmaceuticals, cosmetics, agrochemicals, and new materials. Our mission is to contribute to the advancement of civilization through AI-driven innovations.


[Notes]

*1 SG4D10B

The world's largest molecule-specialized foundation AI model developed by our company. It solves the challenges of conventional molecular AI technology—low prediction accuracy and generalization performance—enabling highly accurate model development even from small-scale datasets held by pharmaceutical companies and chemical manufacturers.


*2 CEATEC 2025

https://www.ceatec.com/ja/application/


*3 Development Achievements of Foundation Model Development Companies in GENIAC

https://www.ceatec.com/ja/conference/detail.html?id=3171


*4 GENIAC

A project aimed at strengthening domestic generative AI development capabilities, termed "GENIAC (Generative AI Accelerator Challenge)," which provides computational resource support for developing foundation models, the core technology of generative AI.

https://www.meti.go.jp/policy/mono_info_service/geniac/index.html


*5 Japanese Companies' Deployment Potential and Overseas Opportunities

https://www.ceatec.com/ja/conference/detail.html?id=3284

Previous
Next