What's in a GGUF, besides the weights – and what's still missing?
Summary
The NobodyWho post analyzes GGUF as a single-file metadata container for language models and surveys what it currently contains (chat templates, special tokens, sampler configuration, and sampling sequence) and what remains missing (tool-call formats, think tokens, projection models, and feature flags). It also discusses how different templating implementations affect performance and interoperability, and suggests ideas for expanding the standard with grammar-based tool calls and improved model-agnostic tooling.