Orb
The world's fastest and most accurate AI model for advanced materials simulation.
Executive Summary
Orb is Orbital Materials' open-source AI simulation model designed to accelerate development efforts in advanced materials. It is recognized as the world's fastest and most accurate model, outperforming proprietary solutions from major labs like Google and Microsoft in both precision and speed, achieving up to 5x faster simulations for large-scale applications. Built on advanced architectures such as Orb-v3, the model leverages innovative techniques like sparse atom graphs to significantly reduce computational cost, making it 10x faster and 5x more memory efficient than alternatives like MACE. Its capabilities have been extended to molecular systems through OrbMol, which utilizes extensive datasets of high-accuracy DFT calculations. Orb is released under a permissive open-source license, allowing free use for non-commercial purposes and startups, thereby maximizing its impact across the materials science community. It is also becoming generally available to AWS customers via Amazon SageMaker, facilitating its integration into cloud-based research and development workflows.
Use Cases
- Simulating advanced materials at scale for research and development
- Accelerating the discovery and design of new materials
- Predicting material properties and behaviors under various conditions
- Developing technologies for data center decarbonization and efficiency
- Extending atomistic simulations to complex molecular systems
Features
Intelligence
- State-of-the-Art Accuracy: Outperforms leading proprietary models from Google and Microsoft in simulation accuracy for advanced materials.
- High-Speed Simulation: Achieves significantly faster simulation speeds, up to 5x faster for large-scale simulations and 10x faster with Orb-v3-direct-20.
- Computational Efficiency: Utilizes sparse atom graphs for 5x more memory efficiency, significantly reducing computational cost.
- Molecular System Extension: OrbMol extends the Orb-v3 architecture to accurately simulate molecular systems using extensive DFT datasets.
Technical Specifications
- Architecture
- Orb-v3 architecture, utilizing sparse atom graphs and an AI-based universal interatomic potential for efficient and accurate atomistic and mesoscale simulations. Leverages the Open Molecules 2025 (OMol25) dataset for training.
- Deployment
- Self-Hosted, SaaS (via AWS SageMaker)
- API Available
- Yes
Infrastructure
- AWS
AI/ML Stack
- AI-based interatomic potentials
- Sparse atom graphs
- Machine Learning
Integrations
- Amazon SageMaker
Pricing
- Model
- Open-source (free for non-commercial/startups), Commercial licensing available
- Starting Price
- Contact sales
- Target Customer
- SMB,Mid-Market,Enterprise,Research Institutions
- Free Trial
- Yes, Indefinite for non-commercial/startups (no credit card required)
About Orbital Materials
Orbital Materials is an AI-first materials engineering company that leverages proprietary AI to accelerate the discovery, testing, and deployment of advanced materials and climate technologies. They develop hardware from the atoms up, focusing on solutions for data centers, including sustainable fuels, carbon capture, and harmful chemical removal. The company's platform solves the inverse design problem autonomously, acting as a generative AI materials engineer.