Alyss like Wonderland
10.5K posts

Alyss like Wonderland
@PreciselyAlyss
GPU-X Cloud DX with startups @ a 🍪 maker 🧚♂️ River Tam energy ✨ EDS & ADHD 💻 opinions expressed are my own ex: @GitHub @Atlassian


Nemotron 3 Super (~4X bigger than Nano) and Ultra (~16X bigger than Nano) are pretrained using NVFP4, a new "Latent Mixture of Experts" architecture that allows us to use 4X more experts for the same inference cost, and Multi-Token Prediction.









sometimes the test prompts can get a little weird lol needed some long ones to test barging. llama 8b can reliably recite the gettysburg address from its weights, in case anyone was wondering 😅




Because AI sucks at writing systems programming code and he works on the linux kernel











We’re announcing a content and product partnership with Vox Media.









