GGUF
GGUF is the single-file format local AI runtimes load: a model's weights, usually quantized, packed together with everything needed to run it.
When a model is prepared for local use, its weights and its tokenizer are packed into one .gguf file. Runtimes like Ollama, LM Studio and llama.cpp open that one file and are ready to chat — no assembly, no separate configuration downloads.
The format comes from the llama.cpp project as the successor to its older GGML files, and it exists because original model releases are not built for laptops: publishers ship full-precision weights split across many large files, meant for data-center GPUs. A GGUF is the compact, usually quantized repack of those weights that fits consumer hardware.
Practically, the GGUF file IS the download that has to fit in your memory. Its byte size is not an estimate — it is a fact you can check before downloading, which is exactly what this site does.
Where you see it on the radar
Every memory figure in our fit tables is measured from the byte size of a real, published GGUF file — the GGUF link next to each model points at the exact file that number came from.
See which GGUFs fit your machine →Related terms