Skip to main content
>_ supraj.dev
TOPIC AI
Local LLM Inference running open-source models on your developer machine
IN 10 SECONDS

Running AI models entirely on your own hardware without sending data to third-party APIs. You download quantized model weights (like Llama 3) and run them locally using your system's VRAM/RAM.

GOTCHA Large models require massive graphics memory (VRAM). If a model is too big to fit in your GPU, it falls back to your CPU, slowing down generation speed to a crawl.
HOW OFFLINE INFERENCE FLOWS
01 You run a command requesting a model like 'Llama 3'.
02 CLI helper downloads the model parameters and loads them into your computer's graphics memory.
03 Local Server starts a private API endpoint running on your localhost.
04 Client sends prompts and receives responses offline with complete data privacy.
POKE IT YOURSELF
ollama run llama3 — pull and start a local chat session with Llama 3
ollama show --system llama3 — inspect the default system prompt template