May 18 · 15 min read · The Problem: One Token at a Time When you ask a large language model a question, it does not write the whole answer in a single step. It generates one small piece of text at a time. That piece is call
Join discussionMay 6 · 8 min read · Got your hands on an NVIDIA DGX Spark but have no idea which models to run on it? I scoured every GB10-optimized model on Hugging Face so you don't have to. Table of Contents What Makes DGX Spark Spe
LLkzwieder commentedMay 6 · 8 min read · An 870 MB drafter model turned Dense 31B from 6.5 → 18.8 tok/s. No model swap, no training, no quality degradation. If you have a DGX Spark, there's no reason not to use this. Key Results Model Fra
Join discussionMay 6 · 10 min read · TL;DR — I was happily running Qwen3.6 on llama.cpp. Then I saw claims of 2× speed with vLLM + NVFP4 + DFlash. So I installed it, fought through crashes, and measured it myself. Verdict: it's real. 88–
Join discussionApr 7 · 10 min read · I recently got my hands on an NVIDIA DGX Spark, and the first thing I wanted to figure out was: how do I access this thing from anywhere? Whether I'm at a coffee shop, at a conference, or on a differe
Join discussion
Mar 14 · 15 min read · There’s a moment when you’re watching a model load into memory. The progress bar is filling up to 87 gigabytes and it hits you. You’re about to talk to something that has 120 billion parameters. Not t
Join discussion
Mar 14 · 58 min read · This guide is for developers only. If you are not yet a Python and Git expert - please learn both before implementing the steps in this article! Python and Git expertise is necessary to run OpenClaw s
Join discussion
