Key Highlights / Executive Takeaways:
• Merck and Nvidia introduce KERMT, a next-generation small-molecule AI model trained on 11+ million compounds to improve early-stage drug discovery decisions;
• Multitask fine-tuning delivers stronger ADMET prediction, cutting preclinical timelines by up to 30% while reducing computational cost;
• Open-source release signals a strategic push toward industry-wide AI acceleration and collaborative drug discovery ecosystems.
KERMT: Advancing Predictive Intelligence in Small-Molecule Discovery: Merck’s KERMT model represents a significant evolution of foundation models for drug discovery, building on the GROVER architecture and fine-tuned for industrial workflows. By improving prediction of absorption, distribution, metabolism, excretion, and toxicity, KERMT helps researchers identify liabilities earlier, reducing late-stage failures that traditionally cost months of experimentation and millions in R&D spend.
Pharma–Tech Convergence at Industrial Scale: The collaboration with Nvidia enables Merck to efficiently scale KERMT across advanced GPU infrastructure, improving speed and memory efficiency compared to previous models. These optimizations not only accelerate molecular simulations but also lower energy consumption, positioning KERMT as a more sustainable, “green” AI model aligned with responsible innovation in pharma R&D.
Real-World Impact on Drug Development Timelines: According to Merck, AI-driven discovery is already compressing early development timelines by 30% or more, improving candidate quality and reducing costs. While clinical trials remain lengthy and high-risk, models like KERMT are strengthening the probability of success by enabling faster target identification and compound optimization, particularly in oncology, cardiovascular, and immunology pipelines.
Open Science and the Future of AI-First Pharma: By releasing KERMT and its genomics-focused counterpart TEDDY as open-source models, Merck is reinforcing a broader shift toward open science and shared innovation. Embedded directly into scientists’ real-time design workflows, these models highlight how multimodal AI—integrating chemistry, genomics, and physics—could become foundational infrastructure for the next era of biopharmaceutical discovery.
About Merck
Merck is a global science and technology company operating at the intersection of healthcare, life science, and electronics. In biopharmaceuticals, Merck is committed to discovering and delivering innovative medicines and vaccines across oncology, cardiovascular disease, immunology, infectious diseases, and more. By integrating advanced data science, artificial intelligence, and high-performance computing into its R&D engine, Merck is accelerating drug discovery, improving success rates, and advancing a more efficient and sustainable model for pharmaceutical innovation.
About NVIDIA
NVIDIA is a world leader in accelerated computing and artificial intelligence, providing the hardware, software, and platforms that power modern AI applications across industries. In life sciences and healthcare, NVIDIA enables large-scale drug discovery, molecular modeling, genomics, and medical research through its GPU-accelerated infrastructure and AI frameworks. By partnering with leading pharmaceutical companies, NVIDIA is helping transform how medicines are discovered, developed, and delivered using scalable, energy-efficient computing.
By