AI hardware in Colombia
GPUs and machines to run AI locally
Graphics cards, NVIDIA DGX Spark, Jetson, big-memory Macs and RTX PCs. Search by VRAM and by the size of model you want to run.
- live listings
- 4 live listings
- cities
- 2 cities
- median price
- $2,3 M median price
Exact finder
What do you want to run?
4 listings match
Like new$4.000.000
Portátil ASUS Zenbook plateado con AMD Ryzen AI 9 HX en venta
Ryzen AI 9 · 32 GB RAM · 1 TB · 16″
Bogotá
Good$660.000
Computador Cpu - Intel I3 7100, 8 Gg Ddr4, 1 Tb Hdd - Nvidia (Usado)
GT 610 · Intel Core i3 · 8 GB RAM · 1 TB
Bogotá
Like newAsk for price
Portátil lenovo loq gamer
RTX 4050 6 GB · Intel Core i5 · 24 GB RAM · 512 GB · 15.6″
Cali
Like newAsk for price
OneXPlayer X1 Pro | Ryzen AI 9 HX 370 | 64 GB RAM | 4 TB | Pantalla 10,95" 2.5K 120 Hz | PC portátil 3 en 1
amd ryzen AI 9 HX 370 · 64 GB RAM · 4 TB · 11″
Bogotá
Which model runs on how much memory?
Rough guide for models quantized to 4 bits with short context. Longer context or higher precision needs more memory.
| Memory (VRAM or unified) | What it runs | Typical hardware |
|---|---|---|
| 8 GB | 7–8B parameter models | RTX 4060, RTX 3070, RTX 5060 |
| 12–16 GB | 12–14B models with room for context | RTX 3060 12 GB, RTX 4060 Ti 16 GB, RTX 5070, RX 7800 XT |
| 24–32 GB | Models up to ~32B | RTX 3090, RTX 4090, RX 7900 XTX, RTX 5090 |
| 48 GB | 70B models | 2× RTX 3090/4090, RTX A6000 |
| 64–128 GB unified | 70B models with headroom, and larger | NVIDIA DGX Spark, Mac Studio or MacBook Pro with 64–128 GB |
Sell yours
What is your GPU or rig worth?
Asking prices from 2 Colombia Move listings over the last year. Listing is free, no commission.
Not enough listings yet for a reference. List yours and be the first.
Price guide from our data
Median and middle-50% range of asking prices across 2 listings from the last year.
Not enough listings yet to show ranges. Every listing makes this guide sharper.
Before you buy a used GPU
- Ask for a video of the card under load (benchmark or stress test).
- Confirm the exact model and VRAM with GPU-Z.
- Check fans, temperatures and that nothing is loose.
- Make sure your power supply and case can take it.
- Pay after testing it in person.
More on the blog: used-laptop checklist · bringing electronics into Colombia (DIAN)
FAQ
How much VRAM do I need to run an AI model locally?
As a rough rule, a model quantized to 4 bits takes a little over half a gigabyte per billion parameters, plus room for context. 8 GB fits 7–8B models, 24 GB fits models up to about 32B, and 48 GB or more fits 70B models.
What is the NVIDIA DGX Spark?
A compact NVIDIA desktop with the GB10 Grace Blackwell chip and 128 GB of unified memory, built to run and fine-tune large models locally. GB10 machines from other brands are similar.
Is a Mac good for local AI?
Yes. Apple-silicon Macs use unified memory, so a Mac with 64 or 128 GB can load models that do not fit on a single consumer graphics card. They are usually slower than a high-end RTX on models that do fit in VRAM.
What should I check when buying a used GPU?
Ask for a video of the card running under load (a benchmark or stress test), confirm the exact model with GPU-Z, check fans and temperatures, make sure your power supply is big enough, and pay after testing it.
How much does it cost to list my graphics card or rig?
It is free with no commission. Pick “Graphics card (GPU)” or “AI hardware” as the device type and fill in the VRAM, and it shows up in this hub and in the search filters.
