VRAMGlass editorial team

VRAMGlass is run by a small team that has been buying used GPUs and running open-weight models on them since 2023, on hardware ranging from a single RTX 3090 to a four-card workstation and a Mac Studio. We write the guides ourselves; AI tools help us gather material and draft tables, and every judgment in a published guide is one we stand behind.

Specs are copied from vendor spec sheets and checked by hand, always as dense (non-sparsity) figures. Used prices are medians of live listings on Xianyu and eBay, refreshed every 6 hours, with outliers and bundles filtered out. VRAM and speed figures are computed from public coefficients on the methodology page, so anyone can recompute them.

Every guide answers one concrete question, contains at least one table that renders live from our database, and states a recommendation with the reason behind it. A person reads the draft, checks each number against its source, and signs off before it goes live. When the data changes we update the text and the modified date; we do not touch dates to look fresh.

If a spec, price rule or conclusion is wrong, tell us and we will fix it and note the change. Guides are limited in number on purpose: a few that stay right beat many that drift.

Reach the team through the contact section of the about page, or check the methodology page for definitions first. Contact · Methodology