Skip to content
AI Stack Nest

Hardware

Everything to pick, install and run open models on your own computer.

Hardware

What you can run depends mostly on memory: how much RAM you have, and whether your graphics card has enough video memory to hold the model. This page lists common RAM sizes with the kind of model each one suits, along with cloud GPU options, so you can decide whether to use the computer you have or rent a GPU by the hour.

Hardware guide
ComputerLargest 4-bit model (estimate)Good for
8 GB RAMMistral 7B Instruct v0.3Learning and small models (about 3B parameters or less)
16 GB RAMGemma 2 9B7B to 8B models for chat and coding help, for example Llama 3.1 8B (about 4.9 GB) or Mistral 7B (about 4.4 GB)
32 GB RAMGemma 2 9BMid-size models around 14B to 27B, for example Qwen2.5 14B (9 GB) or Gemma 2 27B (16 GB)
64 GB RAMLlama 3.1 70BLarge models such as Qwen2.5 32B (20 GB) or Llama 3.1 70B (43 GB), with enough free memory left over