Lizard LLM

Lizard LLM documentation

This is the official starting point for installing Lizard LLM on Windows, understanding its native inference engines, managing GGUF models, and connecting software to its local OpenAI-compatible API.

Install on Windows

Download the current Windows installer, sign in, complete the one-time activation, and let Lizard scan the CPU, RAM, and graphics hardware. The application recommends models that fit the machine and verifies downloads before activation.

Choose and run an engine

Lizard Native uses Direct3D 12 for resident-GPU inference. Caterpillar is designed for CPU and shared or integrated graphics paths. The runtime chooses an engine automatically and can use its bundled llama.cpp fallback or an optional Ollama installation when appropriate.

Connect through the local API

Compatible applications can use Lizard through an OpenAI-style endpoint on the same Windows machine. Change the base URL in the client to the local Lizard endpoint and select an activated model; prompts and outputs remain local.

Use the benchmark CLI and HTTP guide for copy-ready Lizard Native and Caterpillar commands, adaptive concurrency, same-weight llama.cpp controls, and metric interpretation. Read the interactive documentation for other examples, or start with the current Windows download. The FAQ covers privacy, account, payment, GPU, and offline-use questions.