3 ms·

If you’re in the weird position of knowing more about the national labs than the AI lab scene (like I am), link to a TechCrunch profile: https://techcrunch.com

If you’re in the weird position of knowing more about the national labs than the AI lab scene (like I am), link to a TechCrunch profile: https://techcrunch.com/2026/01/28/tiny-startup-arcee-ai-built-a-400b-open-source-llm-from-scratch-to-best-metas-llama/ https://techcrunch.com/2026/01/28/tiny-startup-arcee-ai-buil... My question is: why is this being run out of Argonne? Why not NERSC proper?

frumiousirc 2mo ago
Your question piqued my curiosity. I thought maybe NERSC doesn't accept jobs from private corporations but I checked and that's not true, as long as results are not held proprietary. Perhaps ANL was used as they have a lot of compute and they lead and host the Genesis Open Models Initiative? From your link, 2048 B300 GPUs were used for 6 months. If google search is right, NERSC has 7168 A100. B300's are way more capable than A100. To do this training in 6 months, "3 NERSCs" would be needed. Between the political angle and the technical, I'd guess these two make up a big chunk of the answer.