← Studio Toriumi

Local LLM · GPU · Shops in Japan · 2026-09

Run your own AI at home,
on your own PC.

The most reliable way to own an "uncensored" AI is to run the model on a PC in your own home. Nothing passes through a cloud provider, so there are no provider filters, no usage logs on someone else's server, and no per-token billing. What you need instead is VRAM. This guide covers three things: ① how much memory you need for which models, ② what to check when buying used, and ③ where to actually buy or have a machine assembled around Yokohama, Japan.

See also our companion article on calling your home-PC AI from smart glasses (Rokid Glasses and others) → Smart Glasses Comparison 2026.

We also built a tool that searches used RTX 3090 PCs across Japanese marketplace listings and judges whether the asking price is below market → Used RTX 3090 PC search & value check (in Japanese, with September 2026 sold-price data).

Last updated 2026-09-23 · Sources listed at the end · No hands-on GPU benchmarking was done (prices and specs are from public information)

The short answer

The best first step for most people is a used machine with an RTX 3090 (24GB). As of September 2026 they trade around ¥200,000 in Japan (the Yahoo! Auctions median, up from the ¥130k–180k reported over the summer), run 32B-class models, and the community consensus is still "best value for local LLMs in 2026." With a bigger budget, a new RTX 5090 (32GB); if 16GB is enough for your models, an RTX 5080 or 5070 Ti. Around Yokohama: prebuilt GPU machines at DOSPARA Yokohama and PC Kobo (Pass Kumiai); used bases at Janpara Yokohama and Sofmap Yokohama Vivre; assembly services at Kimura Computer (Aobadai).

Affiliate disclosure: Some links in this article are Amazon Associates (amazon.co.jp) links. If you buy through them, this site earns a referral fee, but the price you pay is unchanged. Shop information comes from official pages and press coverage; inclusion is not affected by link presence.

00Method (evidence disclosure)

How we reviewed this

Criteria: ① usefulness for local LLMs (largest runnable model = VRAM) → ② price/performance → ③ failure resistance (warranty, in-store inspection) → ④ realistic purchase routes in Japan.

Evidence: NVIDIA's official specifications, Japanese technical articles (Zenn/Qiita build write-ups from 2025–2026), market reporting on used prices (as of August 2026), and official shop pages or press coverage (checked September 2026).

Hands-on testing: none claimed. We did not benchmark these GPUs ourselves; speed and price figures are synthesized from public sources.

Updates: First published 2026-09-23 (English). GPU prices move fast — check current listings before buying.

01First principle: for local LLMs, VRAM capacity is king

What limits a local LLM, in order: ① VRAM capacity → ② memory bandwidth → ③ compute. If the capacity is too small the model will not fit at all; if it barely fits but bandwidth is narrow, generation crawls. This ordering is a standard observation in hardware-selection write-ups (e.g. Qiita's "choose by capacity, bandwidth, MoE and TTFT").

Rough guide (assuming 4-bit quantization):

VRAMTypical modelsExperience
8GB7–8B classEntry level. Fast but not bright
16GB13–14B, up to ~20B with effortPractical minimum. The 5080/5070 Ti wall
24GB20–32B classWhere models start feeling smart. Used-3090 territory
32GB32B comfortable; 70B at Q4 barelyEven a 5090 cannot run 70B at full precision (Qiita 2026-04)

Because smarter models consume more memory, the first decision is the model size you want to run, not your budget. Conversely, 32GB is overkill if 8B covers your use case.

02Choosing a GPU in September 2026

Best value · our pick

Used GeForce RTX 3090 24GB

Used ~¥200,000 (Sept 2026 Yahoo! Auctions median, ¥185k–210k)

Still widely called the best-value card for local LLMs in 2026 (xda-developers 2026-03, localaimaster 2026-06). 24GB fits 32B-class models; performance is roughly RTX 5070-class. Two of them (48GB) can target 70B-class models.

Downsides: Used market only. You must judge mining history and epoxy aging (see jisaku.com inspection guides). Warranty is shop-limited. Power draw and heat are high; an 850W+ PSU is recommended.

Maximum speed

GeForce RTX 5090 32GB

New and expensive (still scarce in 2026)

32GB VRAM and top-tier bandwidth make it fastest within its model range. There are documented NVFP4 + TensorRT-LLM speedups (Zenn). For people who can solve budget, a ~1000W PSU and heat exhaust all at once.

Downsides: The price is simply very high (Zenn build articles). Even 32GB cannot hold a 70B at full precision; it only fits at Q4, and inference runs at less than a fifth of 7B speed (Qiita 2026-04). Some SKUs are ~55cm long, complicating case choice.

If 16GB is enough

RTX 5080 / 5070 Ti 16GB

New, mid-high range

Fine up to 13–14B-class models. The CUDA-core difference between the 5070 Ti and 5080 is about 15% against a price gap of roughly ¥100,000, so within 16GB it is largely preference (Zenn).

Downsides: The 16GB wall is hard. If you later want 32B-class models, there is no path except replacement — weaker future-proofing than a used 3090.

03Building the machine — if you buy used

The cheapest route is a used prebuilt tower (12th/13th-gen Core i7 or Ryzen 7, etc.) with the GPU and PSU swapped. Four things to check at a used-PC shop:

Inside used GPUs

Cards previously used for mining may have worn fan bearings and heat-degraded components. For shop stock, choose items with an operational warranty, preferably heavy triple-fan models. Popular models with replacement fans still available (EVGA/ASUS/MSI 3090s, etc.) remain repairable later.

04Buying or assembling around Yokohama — shop guide

All around Yokohama Station and within the city. Presence and stock confirmed from official pages and press coverage as of September 2026. Call or check the web for hours and stock before visiting. (This section is Japan-local by nature; prices in yen, links in Japanese.)

ShopLocationRoleNotes
Janpara Yokohama2 min from Yokohama Station west (Sotetsu) exit
(Nishi-ku Minamisaiwai 1-5-39)
Used PCs, GPUs and partsMajor used-gear chain with shop warranty. A rare place to inspect used GPUs in person
Bic Camera Outlet × Sofmap Yokohama VivreYokohama Vivre 7F (station west exit)Used / outlet PCsSolid warranty coverage. One of the nearest options for a used-PC base
DOSPARA Yokohama EkimaeNear Yokohama Station
(Nishi-ku Minamisaiwai 1-5-30)
Prebuilts (GALLERIA etc.), parts, usedRenewed with stronger parts/used sections (reported by Window's Forest). Demo gaming PCs on display. The place to just buy a GPU-equipped machine
PC Kobo (Unitcom BTO + support)Stores near Yokohama + web BTOBTO and upgrade installation1-year free warranty on BTO (parts + labor). Paid GPU installation/upgrade work accepted through official support
Kimura Computer Yokohama (Aobadai)AobadaiAssembly service, case swapsA specialist that assembles customer-supplied parts. For "I won't build it myself, but I choose the parts"
PC Val (PC Repair 24) Yokohama KannaiKannaiUsed PCs, repairOne station away. An option for bargain hunting

Three buying patterns. A: order a GPU-equipped prebuilt from DOSPARA / PC Kobo (safe, more expensive). B: buy a used tower at Janpara / Sofmap and install the GPU and PSU yourself (cheap, more work). C: gather parts and have Kimura Computer assemble them (the middle ground). For a first build, A or C. B suits people who understand warranties and PSUs.

05Purchase links (affiliates)

Links from this section are Amazon Associates (amazon.co.jp) links. Buying through them earns this site a referral fee at no change to your price.

06FAQ

What does "uncensored AI" mean here?
In this article it means "your own LLM running on your home PC, without a cloud AI's filters, usage logs, or metered billing." With Ollama or LM Studio, your family's text never leaves the house. What you use it for is your own judgment within the law.
Isn't a Mac good enough?
It's not a bad option. Apple Silicon's unified memory (36–128GB) exceeds these GPUs in capacity, and the ecosystem is practical. This guide simply focuses on Windows/Linux towers that can be bought and upgraded at Japanese shops.
Do two 3090s really run 70B models?
There are many reports of Q4-quantized 70B running on 48GB (3090 ×2). But you must plan the motherboard (PCIe slot spacing), PSU (~1200W) and case fit together, so we do not recommend it as a first build.
Can a Yokohama shop just install a GPU for me?
PC Kobo's official upgrade support and Kimura Computer (Aobadai) assembly service accept installation of parts you bring or buy. DOSPARA stores mainly handle their own prebuilts, so call ahead for outside purchases.

07Sources

  1. NVIDIA official — GeForce RTX 5090/5080/5070 Ti and RTX 3090 specifications (VRAM, TGP). Checked 2026-09.
  2. Zenn, "What I considered when building a PC to run LLMs at home" (2025-01, Japanese) — first-hand reasoning on 5080/5070 Ti selection. Checked 2026-09.
  3. Qiita, "Choosing local-LLM hardware by capacity, bandwidth, MoE and TTFT"; Qiita, "Local LLMs on RTX 5090" (2026-04, Japanese) — no full-precision 70B even at 32GB; one-fifth speed at Q4. Checked 2026-09.
  4. xda-developers (2026-03), localaimaster (2026-06), promptquorum (2026-08) — used RTX 3090 assessments and Japanese market prices ¥130k–180k (summer 2026). September prices: median ¥200,000 across 42 Yahoo! Auctions sales (collected for the used local-LLM machine finder, 2026-09-23).
  5. jisaku.com — aging and inspection of used GPUs. Checked 2026-09.
  6. Shops — Janpara Yokohama (Nishi-ku Minamisaiwai 1-5-39), Bic Camera Outlet × Sofmap Yokohama Vivre, DOSPARA Yokohama Ekimae (045-410-0506; renewal reported by Window's Forest), PC Kobo (pc-koubou.jp / pc-support.unitcom.co.jp), Kimura Computer Yokohama Aobadai (045-508-9592), PC Val Yokohama Kannai. Official pages and press checked 2026-09; business conditions change, so confirm before visiting.