AI & Local LLMs

Local models, AI rigs and the tools that run them

Two graphics cards side by side, illustrating GPU VRAM tier comparison

The best local LLM in 2026 isn't one model — it's whichever one fits your VRAM

Llama 4 Scout, Qwen3.6-27B, GPT-OSS-20B, Phi-4-14B — four very different 'best' models, because the real question was never which is smartest. It's what fits.

Lena Fischer · 4h ago · 6 min

Kimi K3 is a 2.8-trillion-parameter AI model. You still can't run it — even with 8 GPUs

Moonshot AI's new open-weight model is bigger than any AI system ever released. The size that makes it good at coding is also the size that makes it unrunnable on anything you own.

Lena Fischer · 5h ago · 5 min

China now processes 140 trillion AI tokens a day. Here's what that actually requires

China's daily AI token usage went from 100 billion to 140 trillion in about two years — a thousandfold jump. ByteDance's Doubao alone accounts for most of it. The number is abstract until you look at what's actually serving it.

Lena Fischer · 9h ago · 6 min