Zatom-1: A Multimodal Flow Foundation Model for 3D Molecules and Materials

2026-02-24

AI Summary

This work proposes Zatom-1, the first unified foundation model capable of jointly generating and predicting both 3D molecules and materials. Addressing the limitation of existing AI approaches that are typically confined to a single chemical domain (molecules or materials) and a single task (generation or prediction), Zatom-1 leverages a Transformer architecture to enable cross-domain representation sharing and transfer. It employs multimodal flow matching to jointly model discrete atomic types and continuous geometric structures, with a unified generative pretraining strategy that provides a general-purpose initialization for diverse downstream tasks. The model achieves state-of-the-art or competitive performance across multiple benchmarks, demonstrates over an order-of-magnitude speedup in generation, and validates positive transfer from material pretraining to molecular property prediction, significantly enhancing both generalization and inference efficiency.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Natural Language Processing: Generation
• Computer Vision: Diffusion Models for Vision

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

General-purpose 3D chemical modeling encompasses molecules and materials, requiring both generative and predictive capabilities. However, most existing AI approaches are optimized for a single domain (molecules or materials) and a single task (generation or prediction), which limits representation sharing and transfer. We introduce Zatom-1, the first foundation model that unifies generative and predictive learning of 3D molecules and materials. Zatom-1 is a Transformer trained with a multimodal flow matching objective that jointly models discrete atom types and continuous 3D geometries. This approach supports scalable pretraining with predictable gains as model capacity increases, while enabling fast and stable sampling. We use joint generative pretraining as a universal initialization for downstream multi-task prediction of properties, energies, and forces. Empirically, Zatom-1 matches or outperforms specialized baselines on both generative and predictive benchmarks, while reducing the generative inference time by more than an order of magnitude. Our experiments demonstrate positive predictive transfer between chemical domains from joint generative pretraining: modeling materials during pretraining improves molecular property prediction accuracy.

Problem

• 3D chemical modeling
• foundation model
• multimodal learning
• generative and predictive tasks
• molecules and materials

Innovation

• foundation model
• multimodal flow matching
• 3D molecular modeling
• generative and predictive learning
• cross-domain transfer