local llm

5 articles · newest first

A home workstation setup — building a local-AI machine in 2026

should you build a local-AI rig in 2026? my honest answer is 'probably not' — with one big exception

this is the question I get more than any other, and my answer disappoints people. the hardware is genuinely capable now. the maths usually still doesn't work — unless one specific thing is true for you.

Lena Fischer · 14h ago · 9 min

Framework Desktop mini-PC, powered by the AMD Ryzen AI Max+ 395 (Strix Halo)

Hardware

a $1,999 mini-PC runs a 120-billion AI model at home. ask it to run a 70B and it chokes

the Framework Desktop and its Ryzen AI Max+ 395 twins hand you ~96GB of GPU memory for two grand — enough to load models no single consumer GPU can touch. whether they actually run fast comes down to one word: bandwidth.

Ravi Malhotra · 2d ago · 7 min

Ollama, the local-model runtime used to run open-weight LLMs on consumer GPUs

a free model on your 24GB GPU codes within 4 points of Claude Opus. meet Qwen3.6-27B

on a single 24GB card in 2026 there's no one 'best' — it's a clean three-way split: Qwen3.6-27B for coding, Gemma 4 for vision, gpt-oss-20b for raw speed. and a used RTX 3090 still competes, because this is a memory game, not a compute one.

Lena Fischer · 2d ago · 7 min

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

everyone asks which local AI model is 'best.' wrong question — your VRAM already picked it

Llama 4 Scout, Qwen3.6-27B, GPT-OSS-20B, Phi-4-14B — four totally different 'best' models, and which one is yours comes down to a single number: how much VRAM your GPU actually has.

Lena Fischer · 2d ago · 6 min

China just dropped an AI model that beats Claude. good luck running it — 8 GPUs isn't even close

Moonshot's new open-weight model is the biggest ever made, and it's genuinely good. it's also flat-out impossible to run at home — and the reason why is a lesson in how GPUs actually work.

Lena Fischer · 2d ago · 5 min