A home for local AI hardware Your next model.
Your next model.
Your own machine.
Find the graphics card or compact computer for your setup. Compare the listings. Get back to building.
Tracking 19 products · Latest price check · USD
Meet your next machine.
All 19 productsCompare the lowest in-stock prices from our recent checks. Photos illustrate the model family; exact boards and configurations vary by listing.
6 featured products · new listings
Apple Mac Studio (M4 Max, 128GB)
AI desktop
- Unified memory
- 128 GB
- Memory bandwidth
- 546 GB/s
No recent in-stock new listing. Check the product page for details.
See product & listing detailsWe may earn a commission from retailer links. Confirm the configuration, condition, shipping, and final price at the retailer. How affiliate links work
The hardware is only half the story.
Figure out memory, quantization, and the setup before you spend.
All setup guidesBest memory value
Compare the catalogLowest recorded dollars per gigabyte. Catalog prices may include used hardware; open a product for its current retailer offers.
From the price watch.
All newsOne Strix Halo box for AI and your NAS? Test both together
Halogen's memory-pressure warnings highlight a buying tradeoff: a box that runs your model may not leave room for household services. Here is what to test.
llama.cpp's 42x drafting result is not a 42x GPU upgrade
New prompt-lookup work speeds up one part of inference. What it measures, what it does not, and how to test your local AI setup before buying more hardware.
RTX PRO 6000 Blackwell drops back to $15,929.99
Today's refresh confirmed the RTX PRO 6000 Blackwell Workstation Edition at $15,929.99 on Amazon, down 11% from yesterday's check.
Ollama flips Apple Silicon to MLX by default in 0.40.0 rc0
Ollama 0.40.0-rc0 now runs supported models on MLX by default on Apple Silicon, which makes today's Mac mini and Mac Studio buying path simpler.