About Cairn — a local AI field guide

What is this local AI field guide?

Cairn tells you which open-weight LLMs your machine can actually run. It reads your GPU through browser APIs, estimates VRAM usage, and guesses tokens per second across 50+ models.

How the VRAM readout works

Each model shows the VRAM it needs at Q4_K_M quantization, in both GB and percent of your VRAM. Over 100% and it won't fit. Tokens per second comes from your GPU's memory bandwidth divided by model size — a rough estimate, usually within 20% of real-world numbers.

Privacy — everything runs in your browser

No data leaves your browser. Detection, scoring, and ranking all run client-side.

Data sources for local LLM specs

Model specs come from the llama.cpp, Ollama, and LM Studio compatibility lists.