Genesis Open Models
The U.S. Department of Energy, in collaboration with industry partners, is advancing a new class of open-weight foundation modelsOpen Model: an artificial intelligence (AI) system released with publicly accessible components—such as model weights, code, or training data—that let developers inspect, modify, and run the software on their own hardware subject to the model's license and other restrictions. specifically to accelerate scientific discovery as part of the Genesis Mission.
These models aim to provide researchers, national laboratories, industry partners, and the broader open scienceOpen Science: practices intended to make scientific processes, outputs, data, software, and methods more accessible, transparent, and reusable, subject to legitimate restrictions. community with powerful, transparent, and extensible AI capabilities that can be adapted across scientific domains and use cases.
The first model in this class is Genesis-Science-1, developed in partnership with Arcee.
Building Shared
Infrastructure for Science
By releasing models with open weights, the U.S. Department of Energy (DOE) seeks to bring the scientific and AI communities together around shared infrastructure for science, enabling new workflows in materials discovery, energy systems, earth systems modeling, fusion biology, high-energy physics, and beyond.
The effort is grounded in open science, reproducibility, and responsible AI, with the goal of lowering barriers to advanced AI capabilities for the public good.
Organizations and researchers can participate by contributing models and data, co-developing evaluations and benchmarks, or collaborating on fine-tuning and application deployments.
Contribute to Genesis Open Models
Genesis Open Models is seeking scientific and technical materials for several stages of model development.
- 01
Foundation-Stage Contributions
Contributions may include scientific text, code, documentation, and structured technical collections suitable for pretraining, midtraining, or context extension.
- 02
Post-Training Data and Environments
Contributions may include supervised fine-tuning examples, expert demonstrations, annotated task examples, research software, datasets, complete workflow environments, reinforcement-learning tasks, held-out evaluations, scoring rubrics, tests, and verifiers.
- 03
Open-Weight Models
Organizations may contribute open-weight models as base models for downstream fine-tuning or immediate deployment. Models should have transparent provenance.
- 04
Fine-Tuning and Applications
Teams can contribute to downstream, domain-adapted versions of Genesis-Science-1 and future models for specific scientific fields, applications, or mission needs, including lab assistants, simulation surrogates, and scientific copilots.
- 05
Scientific Expertise and Infrastructure
The program also needs scientists and engineers who can review tasks, model behavior, and complete artifacts. Infrastructure providers may submit rolling inquiries about additional training or evaluation capacity.
Organizations may apply to both the foundation-stage and post-training contribution tracks when their proposed contributions are distinct.
Additional contributions will be solicited on a rolling basis.
How to Apply
What to Expect When You Apply
Applicants identify the relevant contribution track, describe the scientific work and its importance, explain what materials are ready, name experts available to support the contribution, and state proposed terms of use.
The initial application collects descriptions and metadata only. Scientific materials are not transferred through the application, and contributors specify how their materials may be handled.
Preparing a Data Contribution
Organizations proposing scientific data will first submit an application describing their proposed contribution. If the contribution is determined to be relevant following an initial review, the contributor will be asked to complete a Genesis Data Card.
Genesis Data Cards provide a standardized template and schema for documenting scientific datasets. They provide structured context that enables people and AI systems to find, access, interpret, cite, and reuse data and, when appropriate, integrate it into AI and machine learning workflows.
Data cards support discovery, access, interoperability, reusability, governed and responsible use, and AI usability.
Resources available to contributors include a standardized template and schema, Field Reference Guide, example completed data cards, and validation tools to confirm that completed cards conform to the schema.
View Data Card Template and Guidance
Review & Collaboration
Applications are reviewed through a DOE-governed process with input from academic, national laboratory, and industry partners across five areas:
- 01Scientific fit
- 02Rights and handling
- 03Expert and evaluation readiness
- 04Technical integration
- 05Final program selection
Selected contributors work with researchers building and evaluating Genesis Open Models. They receive early evaluation access as the system develops and are credited in technical reports and release materials.
The program is particularly interested in scientific workflows that are poorly represented by standard AI evaluations and require direct collaboration with researchers who perform the work.
Genesis-Science-1
Genesis-Science-1 is the first model in this new class of open-weight foundation models for scientific research. Developed in partnership with Arcee, it provides an initial foundation for scientific research, downstream fine-tuning, evaluation, and domain-specific applications.
Partners
Genesis Open Models brings together government, national laboratories, academia, industry, scientific nonprofits, and research organizations to advance open-weight AI for scientific research.
Arcee is the first industry partner supporting the Genesis Open Models effort and the development of Genesis-Science-1.
Genesis Open Models welcomes additional organizations interested in contributing models, data, research environments, evaluations, technical expertise, and infrastructure.
Apply to Genesis
Open Models
Universities, national laboratories, companies, scientific nonprofits, research organizations, scientists, engineers, and infrastructure providers are invited to contribute.
Whether you can provide open-weight models, scientific data, research environments, evaluations, technical expertise, or infrastructure, your contribution can help advance shared AI capabilities for scientific discovery.
