---
title: "Local AI at home: hardware and memory guides"
canonical: "https://consideredgear.com/local-ai/"
description: "Plan a local AI setup: estimate model memory, compare mini PCs and desktop GPUs, and check complete-system costs before buying hardware."
---

# Local AI at home

Start with the model you want to run and the computer you already own. These guides help you check memory, software support and the complete purchase cost before choosing hardware. They distinguish sourced buying research from measured results.

## Choose your next step

- [Estimate model memory](https://consideredgear.com/guides/how-much-vram-local-llm/) — Plan for model weights, context and overhead; distinguish dedicated VRAM from shared system memory.
- [Compare hardware types](https://consideredgear.com/guides/best-way-to-run-ai-at-home-2026/) — Compare mini PCs, unified-memory systems and desktop GPUs against your workload and software requirements.
- [Check the complete purchase cost](https://consideredgear.com/guides/ser8-vs-used-rtx-3090-total-cost/) — Separate an upgrade to a compatible desktop from the cost of a complete system.

## Research a specific machine

- [Beelink SER8](https://consideredgear.com/reviews/beelink-ser8/)
- [Used NVIDIA RTX 3090](https://consideredgear.com/reviews/nvidia-rtx-3090-used/)
- [Apple Mac mini M4](https://consideredgear.com/reviews/mac-mini-m4/)
- [Framework Desktop Max+ 395](https://consideredgear.com/reviews/framework-desktop-395/)

[What AI can my computer run?](https://consideredgear.com/tools/local-ai-fit/) Get a first model to try, a useful test and a next step before spending.

## Latest articles

- [Guide] [Ollama vs LM Studio vs Open WebUI: what should you use?](https://consideredgear.com/guides/ollama-vs-lm-studio-vs-open-webui-2026/) — Choose a local AI setup that fits your computer. Compare Ollama, LM Studio and Open WebUI by privacy, ease of use, hardware, local APIs and upkeep.
- [Guide] [How much VRAM do you need for local LLMs](https://consideredgear.com/guides/how-much-vram-local-llm/) — Work out whether a local LLM fits an 8GB, 12GB, 16GB or 24GB GPU. Real model sizes, context-memory math and an Ollama check before you buy.
- [Guide] [SER8 vs used RTX 3090: count the whole computer](https://consideredgear.com/guides/ser8-vs-used-rtx-3090-total-cost/) — Compare a SER8 8845HS with 32 GB/1 TB against a used RTX 3090 in an existing compatible desktop or a new build. A dated research guide, not a benchmark.
- [Blog] [What to check before running Cloudflare Clef at home](https://consideredgear.com/blog/cloudflare-clef-at-home/) — Open weights make Clef worth investigating for local automation. The hardware and software deserve a closer look before you buy anything.
- [Blog] [AI classifiers: small decisions, useful local AI](https://consideredgear.com/blog/ai-classifiers-explained/) — What an AI classifier actually does, how it differs from a chatbot or model router, and where small, testable decisions can help in a home AI setup.
- [Blog] [Jev: a practical guide to TypeSafe's decision model](https://consideredgear.com/blog/jev-decision-model-guide/) — Jev makes typed decisions through a hosted API. Here's what its choice, score, and yes/no outputs mean, what it costs, and what to check before using it.
- [Blog] [Laya: an open-source classifier for local decisions](https://consideredgear.com/blog/laya-open-source-classifier-guide/) — Laya brings typed AI decisions to your own hardware. Understand its open license, model choices, hardware needs, and the limits behind the benchmark headlines.
- [Review] [Beelink SER8: a compact local-AI candidate](https://consideredgear.com/reviews/beelink-ser8/) — Sourced buying notes for the SER8 8845HS with 32 GB/1 TB. Check the exact configuration and software support; current price and performance are unverified.
- [Review] [Framework Desktop review: the home AI machine you can actually open](https://consideredgear.com/reviews/framework-desktop-395/) — 128 GB of shared memory in a small box, running large AI models that a normal graphics card can't fit. A premium-priced system with accessible parts, though its memory cannot be upgraded later.
- [Review] [Mac mini (M4) review: the easiest way to run your own AI](https://consideredgear.com/reviews/mac-mini-m4/) — Apple's small desktop turns out to be one of the best-value machines for running AI at home — if you buy the memory right at checkout and accept that you can never change it later.
- [Review] [Used RTX 3090: check the whole-system cost](https://consideredgear.com/reviews/nvidia-rtx-3090-used/) — A used 24 GB card may suit a compatible desktop. Check the exact board, supporting hardware and seller terms before comparing complete-system costs.
- [Blog] [You don't need the cloud to run your own AI](https://consideredgear.com/blog/you-dont-need-the-cloud-for-ai/) — A plain-English explainer on running AI models on your own hardware — what it means, what it costs, what you give up, and why it's the front door to everything else we cover here.