Home Rig
The hardware side of local AI: which GPU, how much RAM, rent-or-buy math, which software runs what, and what quantization really costs you.
The rig pages answer the five practical questions behind every “can my machine run it” search: where each model actually runs, what the parts cost right now, what config a home machine needs, which software to use, and what quantization gives away. Every page carries a dated status board — hardware prices are moving monthly in 2026, and we date every number.
-
Which Software Runs the Big Three? Strata, Maya, and the GGUF World, Compared
The engine is the real 2026 decision: a guided launcher (Strata), a three-tier scheduler (Maya), or classic GGUF tools. What each runs, what it costs, and where each one disappoints.
-
What Quantization Costs You: the IQ1→Q8 Ladder With Measured Numbers
Quantization is how 428B fits in 93 GB — but what does it give back? Real KLD numbers, task-level scores, and why 2026's QAT models changed the ladder.
-
What GPU Do You Need to Run the Big Three at Home? VRAM Tiers, Checked November 2026
Which card runs Qwen3.8-Flash-Next, GLM-5.3-Flash and MiniMax M3.1 — and why 2026 GPU prices are a memory story. Every speed number is a measured one, every price is dated.
-
Rent or Buy? What Running the Big Three Actually Costs — Tokens, Hours, or Silicon
Three price layers checked and dated: per-token API and aggregator rates (Nov 3), hourly GPU rental (late Oct), and what a home rig costs after the 2026 memory squeeze — plus the break-even formula.
-
How Much RAM — and It Costs What Now? The 2026 Memory Squeeze, Checked November
RAM decides whether the Big Three fit. In 2026 it also decides your budget: DDR5 contract +10–15% this quarter, DDR4 declared EOL, spot prices wild. The sizes, the dates, the buy-now-or-wait math.