How CraftRigs Works
What this site is
CraftRigs answers one question: what hardware do you actually need to run AI models locally? VRAM requirements, GPU comparisons, quantization trade-offs, and the troubleshooting that vendor spec sheets skip.
AI-assisted editorial process
CraftRigs is an AI-assisted publication. Our articles are researched, drafted, and updated with the help of AI tools, working from primary sources, and organized under four editorial personas (Georgia Thomas, Ellie Garcia, Chloe Smith, and Charlotte Stewart) that each own a beat. The personas are editorial identities, not individual human reviewers — each one represents a consistent methodology and coverage area, described on its author page.
We disclose this because you deserve to know how the recommendations you read are produced. Every page on this site is reviewed against the sourcing rules below before and after publication, and pages that no longer meet them are updated or removed.
Where our numbers come from
- VRAM and memory requirements — computed from model parameter counts, quantization formats (GGUF/GPTQ/AWQ), and KV-cache math, cross-checked against llama.cpp, vLLM, and Ollama documentation and issue trackers.
- Hardware specs — manufacturer specification sheets (NVIDIA, AMD, Intel, Apple) — never secondary coverage.
- Benchmarks — published results from the organizations that produced them, with the test configuration stated. We do not invent benchmark numbers; where community results vary, we say so.
- Pricing — public retail and used-market listings at the time of writing, with dates. Prices drift; check the date on the page.
Corrections
If a spec, requirement, or claim on this site is wrong, we want to fix it. Use the contact page and include the article URL. Corrected pages show an updated date.
What we don't do
- No sponsored rankings — no manufacturer pays for placement.
- No invented lab results — claims trace to a named public source.
- No affiliate-driven recommendations: when we say skip a product, we mean it.