Tools/Infrastructure

Local AI Model Picker

Tell it your computer's RAM and get the best open model to run privately offline.

Your computer's RAM (or unified memory)16 GBApple Silicon Mac (M-series) — unified memory runs larger models efficiently

How to run it

Install LM Studio (friendly) or Ollama (terminal).
Download a model from the list on the right.
Chat fully offline — nothing you type leaves your machine.

Recommended tier

Solid

Models to try

Llama 3.1 8B, Qwen 3 8B, Gemma 4 9B (quantized)

What you can do

The sweet spot — genuinely useful reasoning, coding help and offline research.

Guidance, not gospel — exact fit depends on quantization (4-bit lets bigger models fit), context length and your GPU. Start one tier below your max for snappy responses, then size up.

Read the Infrastructure dimension All tools