Los Alamos Expands OpenAI Models on Venado-GPT
Los Alamos National Laboratory marks one year of its partnership with OpenAI as use of the models grows across three national laboratories.
Updated on Nov. 21, 2026 in Artificial Intelligence

Los Alamos National Laboratory marked one year of its partnership with OpenAI, bringing three experimental models to its secure computing environment through Venado-GPT. Usage of the service has grown sharply across the tri-laboratories, with weekly token throughput rising 20-fold to more than 3 billion.
Why it matters
The partnership shows how frontier AI models can run inside secure, on-premises high-performance computing environments used for sensitive national security science. Its growth across Los Alamos, Lawrence Livermore and Sandia suggests broader adoption of AI tools in U.S. national laboratory research.
Venado-GPT runs the three OpenAI models entirely on Venado's local graphics processing units, linked by the Ethernet-based HPE Slingshot interconnect. Weekly token throughput grew 20-fold to more than 3 billion, but the source does not specify latency or cost per request.
The players
Los Alamos National Laboratory
A U.S. Department of Energy national laboratory in New Mexico that conducts research in national security, energy and science, and operates the Venado supercomputer.
OpenAI
An artificial intelligence research company that develops the GPT family of large language models, including the models deployed on Venado-GPT.
The details
Venado entered service in 2024 and was designed for scientific workflows. Staff at Los Alamos, Lawrence Livermore and Sandia use the service, where researchers have applied it to streamlining workflow handoffs between scientists and engineers, automating translation of simulation models, and identifying kernel and driver changes during network maintenance.
Timeline
2024: Venado enters service.
Mid-February 2026: Measured growth in Venado-GPT token throughput and usage begins.
Late August 2026: The measured growth period for Venado-GPT usage ends.
Year two of the partnership: Planned enhancements to Venado-GPT are expected.
Artificial Intelligence Adoption Across U.S. Labs
The partnership places Los Alamos, Lawrence Livermore and Sandia, three laboratories within the Department of Energy's national laboratory system, in a shared effort to deploy frontier AI on secure supercomputing hardware. This extends the laboratories' longstanding role in high-performance computing into large language model use.
Researchers across the three national laboratories can now use the same OpenAI models for workflow tasks such as translating simulation models and tracing network changes. Year two of the partnership will bring enhancements to file attachment reading and application programming interface reliability.
The takeaway
Secure, on-premises deployment lets sensitive research teams adopt frontier AI tools without sending data to outside cloud services. The growth in use suggests that this model may be replicated at other scientific institutions.
Further reading
For more on AI deployments in government research, explore Artificial Intelligence coverage on Wisevoter.
Source note: This article includes information reported by Los Alamos Reporter.






