GGUF Metadata Comparator checks tensor quants
This single-file HTML gist lets you compare GGUF model metadata, tensor shapes, and layer quantization levels side by side. It’s a quick way to spot whether different omnicoder quants actually match before you waste time loading or benchmarking the wrong one.
Tiny browser tools like this are exactly the kind of practical glue that saves time in model-hacking workflows. It won’t replace full GGUF inspection suites, but it makes “are these weights really the same, and which tensors changed?” answerable in seconds.
- –Compares a reference GGUF against multiple targets, which is handy when you’re sanity-checking quant variants
- –Surfaces metadata, tensor shape differences, and quantization-level changes in one view
- –Uses the Hugging Face `gguf` parser, so it stays lightweight and easy to share as a gist
- –The “show unchanged rows” toggle makes it easier to focus on real drift instead of noise
- –Especially useful for model tinkerers deciding which layers or weights are safe to quantize
DISCOVERED
70d ago
2026-03-18
PUBLISHED
70d ago
2026-03-17
RELEVANCE
AUTHOR
Interpause