Inspect locally
Explore GGUF metadata, tensors, and ONNX graph details without sending the file to a server.
Local GGUF & ONNX analysis
Private AI model inspection directly in your browser.
Inspect metadata, explore model structure, compare files, and estimate inference RAM. Your model stays on your device.
No account or installation required. The project is public on GitHub .
Before the AI takes over the world, stop and buy me a coffee.
Plan inference
Estimate RAM for different context lengths, parallel sequences, and GGUF quantizations.
Compare models
Load two files of the same format to find shared, changed, and model-specific data.
Download parsed results as JSON or calculate a SHA-256 fingerprint on demand. Everything runs in a recent browser. Found an issue? Report it on GitHub.
What can AI Model Inspector inspect?
For GGUF files, the inspector reads the header, typed metadata, and tensor descriptors. For ONNX files, it reads model properties, graph inputs and outputs, nodes, initializers, and tensor relationships.
Estimate memory before running a model
The estimator separates model weights, KV-cache memory, and a runtime allowance. Change context length and parallel sequences to see how the estimate responds.
Your model stays on your device
Parsing and optional SHA-256 hashing happen in your browser. The model file is not sent to an AI Model Inspector server. Ordinary page requests and advertising are separate; see the privacy policy.