At launch, RHEL AI contains help for the Granite 7-billion-parameter English language mannequin. One other Granite mannequin, the 8-billion-parameter coding mannequin, is in preview and can be usually accessible on the finish of this yr or starting of 2025.
RHEL AI additionally comes with Instruct Lab, an open-source mission that helps enterprises superb tune and customise these Granite fashions, or different open-source AI fashions.
Lastly, RHEL AI additionally comes with all of the underlying platform infrastructure, Katarki says. That features speedy help for Nvidia {hardware}. Assist for AMD and Intel {hardware} is anticipated to reach within the subsequent few weeks.
The whole lot is packaged up as a container picture, Katarki provides, in order that enterprises can use their current container administration instruments to manage it. On high of the software program, there’s additionally help and authorized indemnification for each the open-source software program and the Granite mannequin.
“Consider it as an equipment,” Katarki says. “It’s bought every part. The Granite mannequin, Instruct Lab, all of the underlying platform software program you want, and the working system beneath all of it.”
RHEL AI helps enterprises get away from the “one mannequin to rule all of them” strategy to generative AI, which isn’t solely costly however can lock enterprises right into a single vendor. There at the moment are open-source giant language fashions accessible that rival these accessible from the industrial distributors in efficiency.
