AI Model Inspector documentation

These pages explain the information the inspector reads and how its estimates are calculated. They follow the current source code, including its limits.

File formats

GGUF filesHeaders, metadata, tensor shapes, types, and parser limits.ONNX filesModel properties, interfaces, nodes, initializers, and supported protobuf fields.GGUF metadataArchitecture, context, attention, and tokenizer-related keys.ONNX graphsHow inputs, outputs, nodes, tensors, and dependency traces fit together.

Calculations and workflows

Inference memoryThe exact formulas, defaults, assumptions, and a worked example.GGUF quantizationHow alternative file-size and memory ranges are projected.Compare modelsWhich GGUF and ONNX fields are compared and what a match means.

Using the reference

Start with the page for your file format, then use the focused pages when a field or estimate needs more context. Return to the inspector at any time to check your own file. No model file is needed to read these pages.