Local LLM GPU Value Weekly Report

Stop cross-referencing six browser tabs to figure out which GPU to buy for local LLMs. Every week: live street prices (used and retail) matched against community llama.cpp benchmarks, ranked by what actually matters: dollars per token/second and dollars per GB of VRAM. Every number cited to its source and date. Nothing estimated.The current week's issue is always free. Grab it here and check the quality yourself. Subscribing gets you the working toolkit: The full dataset as CSV, every week — sortable, filterable, with a clickable buy link for every row Week-over-week deltas from your second issue onward — what moved, by how much, so you catch the buy window The complete back-issue archive — price history since you started deciding Every issue emailed on release — no checking back This week (W27): a used RX 7900 XTX at ~$880 is the perf-per-dollar leader at $5.17 per tok/s. The RTX 5090 dropped ~17% week-over-week. Used A6000s are quietly the cheapest 48GB path.Most people subscribe for a month or two while they're shopping, then cancel. That's fine by us. That's what it's for.Every figure cited. Cancel anytime.

Found on Devadex — the discovery index for independent software the big search engines bury. More from gumroad.

Report this listing