Free › Free AI tools › Local model memory calculator
Free live tool
Local model memory calculator.
Will that model run on your machine? Pick a model and a size, and see the memory it needs.
Runs in your browser. Nothing is sent anywhere.
Calculator
How much memory does a local model need?
Includes open models such as Llama and Mistral. Results are estimates.
Your model
Memory needed
Will it fit?
Estimates only. Real use varies with the software you run (such as llama.cpp, Ollama or MLX), the exact model file and other programs open. Figures use decimal gigabytes.
How it works
Three things fill the memory.
Weights.
Parameters multiplied by the bits each one uses. Compression such as 4-bit makes a model several times smaller, with some loss of quality.
Context memory.
The longer the conversation or document, the more memory it uses. Long contexts can need several extra gigabytes.
Overhead.
The software itself and the rest of your computer. On a Mac, only part of the unified memory is available to the model.
Running models locally is a good fit when data must stay on your own hardware. Not sure where to start? AI consulting ›
More free tools.
Lessons, tools and datasheets. No sign-up.