Can my PC run it?
Open AI models can run on your own computer, for free and without sending your data anywhere. Tell us what you have and see which ones fit.
Estimates for a conversation of about 8,000 tokens. Long documents need more memory, and speed also depends on the rest of your computer. Model sizes come from Hugging Face. 4-bit, compressedi: Compression shrinks a model so it fits in less memory: 8-bit is nearly identical to full quality, 4-bit is a little less accurate but takes about a third of the memory.
How to run a model on your computer
- Install a free app such as Ollama or LM Studio (Windows, Mac and Linux). Ollama ↗ · LM Studio ↗
- Search for the model by name in the app and download the compressed version (often labelled Q4_K_M) that fits your memory.
- Chat with it offline: nothing you type leaves your computer.