Insilico Medicine Launches its Multimodal Generative Chemistry Model on AWS Marketplace
Insilico Medicine has released Nach01, a large-scale generative chemistry foundation model, on the AWS Marketplace. The model combines structural and spatial representation learning and is built to handle hundreds of tasks spanning retrosynthesis, molecular prediction, and other generative and predictive chemistry applications.
Nach01 was trained using Amazon SageMaker, AWS’s managed machine learning service, and is available for inference and fine-tuning either via Python or API access. It supports 2D and 3D predictive modeling and can be deployed directly within AWS environments through SageMaker.
The model is the third iteration in Insilico’s Nach0 model family, extending capabilities from earlier versions:
- Nach0, a cross-domain language model for chemistry and natural language tasks.
- Nach0-pc, an architecture that integrates textual and spatial representations to handle atom-level geometry.
- Nach01, which combines both and is optimized for general-purpose chemistry reasoning and molecule generation across hundreds of task types.
According to Insilico, Nach01 is designed to be embedded in experimental workflows, acting as an “oracle” to guide molecule generation, support reinforcement learning pipelines, and increase the success rate of experimental validation. It can be fine-tuned on proprietary datasets and used to generate or predict activity, ADMET properties, and even quantum-level calculations, making it applicable across hit discovery, lead optimization, and molecular property modeling.
CEO Alex Zhavoronkov described Nach01 as "one of the most capable multimodal foundational models in our arsenal and in the industry", citing its role in advancing Insilico’s vision of pharmaceutical AI systems capable of generating novel chemical matter at scale. AWS VP Jon Jones noted that the release reflects AWS’s commitment to making generative AI tools broadly available to accelerate biomedical innovation.
Topics: AI & Digital
