What each machine can run
For every card and computer: the best models for each task, how fast they run, and what it costs. Your machine not here? The checker on the home page knows over 90 chips.
Graphics cards
The fastest single card you can buy, at a very high 2026 price.
Still a top 24 GB card. Runs the best 27–32B models at a smooth speed.
The best value for 24 GB. Six years old, still the community favourite.
48 GB of graphics memory for less than one new 32 GB card. Loud and power-hungry.
Fast, but only 16 GB. Great for models up to about 20B.
Nearly the RTX 5080’s speed for less money. Same 16 GB limit.
NVIDIA’s cheapest new card with 16 GB. Quiet and low power, but slower.
The cheapest useful NVIDIA card. Fine for 8–12B models.
AMD’s fastest current 16 GB card. Good speed; software takes a little more care.
24 GB with RTX 4090-class bandwidth. Best if you already own one.
Cheap 12 GB card. Works for small models, with more setup.
48 GB on one card. Slower than new cards, but fits 70B models.
Macs
A quiet always-on box for small models (up to about 8B).
A quiet desk machine for models up to about 14B, and faster than the M6.
Silent and portable. Fine for small models; slow on big ones.
A laptop that runs 27–35B models. The sweet spot for developers on the move.
The fastest laptop for local AI. Runs 70B-class models.
A silent desk machine for 30–70B models. Good value among Macs.
The most capable Mac for local AI. Runs 70–120B models.
Small boxes with lots of memory
A small box with 128 GB that runs 100B+ mixture-of-experts models.
A small PC with 128 GB shared memory. Great for big mixture-of-experts models.