Argonne Lab Opens AI Inference Service for Scientific Research
The US Department of Energy's Argonne National Laboratory launched an AI inference service that gives researchers access to large language models and specialized AI systems running on federal supercomputers. The service eliminates the need for scientists to build and maintain their own AI infrastructure.
Researchers can now use the Argonne Leadership Computing Facility (ALCF) Inference Service to analyze datasets, test hypotheses, and process experimental results without weeks of hardware and software setup. The platform operates on ALCF systems including Sophia and Metis, with expansion planned to additional NVIDIA-based machines.
How the Service Works
The system supports commercial and open-weight models such as Google Gemma, Meta Llama, and OpenAI's GPT-OSS models, plus in-house systems like AuroraGPT. Authorized researchers from universities, industry partners, and national laboratories including Brookhaven, Lawrence Berkeley, and Oak Ridge can access the platform using their home institution credentials.
The infrastructure uses Globus Compute and Globus Auth technologies to enable federated access and distributed workflows across institutions.
Reducing Research Costs
Advanced AI research workflows often require repeated exchanges between AI models and simulation software-a process called tool calling. These interactions consume large volumes of tokens, making commercial cloud platforms prohibitively expensive for many research teams.
By using a shared federal computing system, research organizations avoid major capital expenditures on graphics processing units, complex software maintenance, and recurring subscription fees to commercial AI vendors. This cost reduction allows institutions to redirect funding toward experiments, materials development, and staffing.
Current Applications
The service is already supporting the DOE Genesis Mission, which uses AI models to analyze experimental data streams and predict plasma disruptions in fusion energy research. In chemistry and materials science, tools like ChemGraph are automating molecular simulation workflows on the platform.
Michael Papka, director of the ALCF, said the service "helps close the gap between developing AI models and putting them to work in scientific research" by offering AI inference as a shared resource.
Learn more about Generative AI and LLM applications, or explore AI for Science & Research training resources.
Your membership also unlocks:
