GGUF Toolkit

Five tools for .gguf artifacts — inspect the header, size every quantisation, validate the self-describing KVs, inventory tensors, build a ComfyUI workflow. Everything runs in your browser: the file is never uploaded, no server is involved, and no CPU is spent by the host. Only the file's header is read, so even a multi-GB model is instant.

Drop a .gguf anywhere on this page, or choose one above. Nothing leaves your machine.

Every metadata KV and the quantisation mix, read from the container only.

Load a .gguf file to begin.

What a model costs in every quantisation, before you download. Bits-per-weight from the ggml type table.

or 
Enter a parameter count, or load a .gguf and compare its real tensor sizes.

Does the file carry the self-describing metadata loaders need to classify it from the header alone?

Load a .gguf file to begin.

Every tensor, shape and quant type — from the index only, no weights read.

Load a .gguf file to begin.

Emit a loadable ComfyUI workflow JSON for an AEUPH GGUF (needs ComfyUI-GGUF). Runs on your machine, not this host.

Build a workflow to see the JSON.