LLM Configurator matches local models to your hardware

LLM Configurator is a free decision engine that helps users choose models, hardware and software for running AI locally. It checks which of 194 published open-weight model variants fit a GPU or Mac, estimates their speed and links to 125 setup guides. Hardware detection runs in the browser without uploading the detected information, and no account is required.

Compatibility calculations account for model weights, context cache and runtime overhead rather than parameter count alone. Models that require spilling into system RAM are listed separately because they run much slower. Recommendations prioritize memory capacity, then bandwidth, with workload-specific options for coding, document search, writing and other tasks. A shared calculation engine keeps results consistent across the tools.

Most performance figures are estimates, clearly distinguished from measurements. The speed model is calibrated against 14 measured runs across 3 GPUs, credited to third-party sources, and presents estimates as ranges rather than promises. Additional tools compare local hardware and electricity costs against cloud services, suggest upgrades and generate builds around budgets and target models. Setup guidance covers Ollama, LM Studio, llama.cpp, vLLM and MLX, although format differences mean some runtime estimates should be treated as guidance.