Free AI Hardware Checker

Run an LLM locally: find out exactly what your computer can handle.

Running an LLM locally on your own machine saves money and keeps your data off the cloud. The catch: not every PC is ready for every model. This free tool matches your hardware to the models that genuinely fit — no guesswork, no wasted downloads.

Three things worth knowing first

VRAM (Video RAM) is the dedicated memory built into your graphics card. It is not the same as your system RAM — it is a separate, faster pool used exclusively by the GPU. When you run an AI model locally, the model’s weights (billions of numbers that encode what the model has learned) need to live in VRAM while it works. A 7-billion-parameter model in its standard form needs roughly 14 GB of VRAM. Most consumer gaming GPUs ship with 8 GB or 12 GB, which is why model compression matters so much.

Quantization is the process of compressing a model by lowering the numerical precision of its weights — from 16-bit floating-point numbers down to 8-bit integers or even 4-bit values. A 4-bit quantized 7B model can fit in as little as 4–5 GB of VRAM with only a small reduction in output quality. This is the single biggest reason consumer hardware became viable for local AI: the same model that once required a $10,000 workstation GPU can now run on a mid-range gaming card you already own.

GGUF is the file format used to distribute quantized models for local inference. Files end in .gguf and are designed to load efficiently in popular local runtimes like Ollama and LM Studio. When you see a model listed as Q4_K_M or Q8_0, those are quantization labels embedded in the GGUF filename — they tell you exactly how compressed the model is and how much VRAM you need before you download a single byte.