- Argonne has launched a new AI inference platform for researchers using advanced AI models
- The inference service provides access to major AI models from Google, Meta and OpenAI
- The DOE’s Genesis Mission and American Science Cloud initiatives will leverage the service
The Department of Energy’s Argonne National Laboratory has launched an artificial intelligence inference service designed to provide researchers with scalable access to advanced AI models running on high-performance computing infrastructure.
Federal agencies are continuing to invest in AI, cloud infrastructure and cybersecurity to support mission modernization and operational resilience. Learn how government and industry leaders are navigating these priorities at the Potomac Officers Club’s 2026 FedCiv Summit on October 29. Save your seat now.
Table of Contents
What Does the ALCF Inference Service Provide?
Argonne said Tuesday the Argonne Leadership Computing Facility, or ALCF, Inference Service offers cloud-like access to large language models, such as Google’s Gemma series, Meta’s LLaMA models and OpenAI’s GPT-OSS family. The service also provides access to domain-specific foundation models, computer vision models and internally developed systems, like AuroraGPT.
How Does the ALCF Service Support Federal Research Programs?
The platform serves as a foundational tool for the DOE’s Genesis Mission, a national initiative to build an advanced scientific platform, and integrates with the American Science Cloud. Its current user base spans multiple institutions using federated home credentials across national facilities, including Oak Ridge, Los Alamos, Lawrence Berkeley and Lawrence Livermore national laboratories. Applications include predicting plasma disruptions in fusion energy experiments, filtering collider data in high-energy physics and deploying agentic frameworks like ChemGraph to automate molecular simulations more efficiently.
The launch builds on broader DOE and Argonne initiatives to expand AI-enabled scientific research infrastructure, including the Genesis Mission funding program announced earlier this year and Argonne’s partnerships with NVIDIA, Fujitsu and Riken to advance AI and high-performance computing capabilities.
How Is the Service Expanding Within DOE?
The inference platform already supports researchers from multiple national laboratories, including Oak Ridge, Los Alamos, Lawrence Livermore, and Brookhaven. Argonne said the service currently runs on the Sophia and Metis systems and will later expand to NVIDIA-based systems, including Tara and Minerva.
