KERMT
AI model · Science / biology model by Merck (MSD)
Advancing drug discovery with open-source AI foundation models.
KERMT is an open-source AI foundation model for small molecule drug discovery that predicts molecular behavior and cuts development timelines by approximately 30%. It is designed for scientists and researchers to optimize drug candidates during the lead optimization phase.
Model specs
The full spec card (context, pricing per token, modalities and more) on AntSpace. Sign up free to see the full spec card.
- Launched
- 2026-03-19
- Languages
- English
Category: Drug Discovery
- Domains
- Small molecule drug discovery, lead optimization
- License
- Open-source
- Access
- Nvidia BioNeMo, Clara Open Models, GitHub
- Released
- 2026-03-19
- Full Name
- Kinetic GROVER Multi-Task
- Training Data
- 11 million molecules
- Trained on 11M+ molecules
- Cuts drug development timelines by ~30%
- Open-source deep-learning foundation model
- Available on Nvidia BioNeMo and Clara Open Models
Overview
KERMT (Kinetic GROVER Multi-Task) is Merck's open-source deep-learning foundation model designed to advance small molecule drug discovery. Developed in collaboration with Nvidia, it is trained on more than 11 million molecules to help scientists predict molecular behavior in the human body. The model is primarily used during the lead optimization phase to fine-tune early molecules, potentially cutting drug development timelines by 30% or more. By identifying disease targets and optimized compounds faster, KERMT reduces the need for extensive lab work and increases the probability of success in preclinical phases. It is available to the scientific community through Nvidia accelerated computing platforms like BioNeMo and Clara Open Models.
Drug Discovery Capabilities
- 11M+ Molecule Training: The model is trained on a dataset of more than 11 million molecules to learn chemical patterns and behavior.
- Molecular Behavior Prediction: KERMT helps scientists better predict how a molecule will behave in the human body during the lead optimization phase.
- Timeline Acceleration: The model identifies disease targets and optimized compounds faster, potentially cutting drug development timelines by 30% or more.
- Early Issue Detection: By using AI/ML predictions, researchers can spot potential issues with drug candidates much earlier and reduce the need for months of lab work.
Access and Collaboration
- Open-Source Code: KERMT is an open-source model with its code available to the entire scientific community.
- Nvidia Platform Integration: The model is available on Nvidia accelerated computing and software platforms.
- BioNeMo Accessibility: KERMT is accessible through platforms such as Nvidia BioNeMo and Clara Open Models.
Use cases
- Scientists use KERMT during lead optimization to predict how a molecule will behave in the human body.
- Drug development teams use KERMT to spot potential issues with candidates early, reducing months of laboratory work.
Specifications
- Domains
- Small molecule drug discovery, lead optimization
- License
- Open-source
- Access
- Nvidia BioNeMo, Clara Open Models, GitHub
- Released
- 2026-03-19
- Full Name
- Kinetic GROVER Multi-Task
- Training Data
- 11 million molecules
- Architecture
- Deep-learning foundation model
- Primary Phase
- Lead optimization
- Collaboration Partner
- Nvidia
How it works
- Access the Model: Researchers access the open-source code or use the model through Nvidia's accelerated computing platforms like BioNeMo.
- Predict Molecular Behavior: Scientists apply the model to their chemical data during the lead optimization phase to predict how molecules will behave.
- Optimize Drug Candidates: The model identifies the most promising drug candidates and potential issues, allowing teams to focus their lab work on optimized compounds.
FAQ
What is KERMT?
KERMT is a deep-learning foundation model trained on more than 11 million molecules to help scientists predict molecular behavior in the human body.
How is KERMT used in drug discovery?
The model is primarily used during the lead optimization phase to fine-tune early molecules and identify optimized compounds faster.
How much time can KERMT save in drug development?
KERMT can potentially cut drug development timelines by 30% or more by accelerating the identification of disease targets and optimized compounds.
Is KERMT open source?
Yes, KERMT is an open-source model, and its code is available to the entire scientific community.
Where can I access KERMT?
KERMT is available on Nvidia accelerated computing and software platforms, including Nvidia BioNeMo and Clara Open Models.
Added by AntSpace from public sources (merck.com).