A year into its OpenAI partnership, LANL reports surging use of AI models on Venado
Weekly token throughput on Venado-GPT grew 20-fold from mid-February to late August, the Lab says.
LOS ALAMOS - A year after Los Alamos National Laboratory began a partnership with OpenAI, use of the company's artificial intelligence models on the Lab's Venado supercomputer has climbed sharply across three national laboratories, according to a Lab news release published Monday.
From mid-February to late August 2026, weekly token throughput on the service grew 20-fold to more than 3 billion tokens, the Lab says. User accounts tripled, and weekly successful requests rose from about 1,200 to more than 41,200.
The service, Venado-GPT, runs three experimental OpenAI models entirely on Venado's own graphics processors and is open to staff at Los Alamos, Lawrence Livermore and Sandia national laboratories. In a nod to J. Robert Oppenheimer, the models are nicknamed 5.4 OppenAI, 5.4 OppenAI mini and o3 OppenAI.
Los Alamos staff have paired the models with URSA, open-source agent software developed at the Lab, to support certification workflows, the release says.
"We've seen tremendous growth in the tri-laboratories' use of Venado-GPT," Director Thom Mason said in the release.
In the partnership's second year, the Lab and OpenAI plan to improve the models' handling of file attachments, application programming interfaces and complex coding work.