# Debugging Tools

The following tools are available for debugging and inspecting the runtime execution of the AI model.

- [Accuracy Debugger](https://docs.qualcomm.com/doc/80-62010-1/topic/accuracy_debugger.html#accuracy-debugger): QAIRT provides the `qairt-accuracy-debugger` tool for identifying inaccuracies in neural networks at the layer level. This tool performs a comparison between reference (golden) outputs generated by executing a model within an ML framework and the corresponding outputs obtained when running the same model on target hardware platforms, such as NPUs, CPUs, or GPUs.
- [QAIRT Optrace](https://docs.qualcomm.com/doc/80-62010-1/topic/qnn-optrace.html#optrace): Optrace profiling gives the exposure to an AI Runtime SDK user on how a AI model/network was executed on the NPU hardware. It provides a detailed breakdown of how the network was executed on the NPU hardware.

Last Published: Jun 16, 2026

[Previous Topic
Quick AI Application Builder](https://docs.qualcomm.com/bundle/publicresource/80-62010-1/topics/QAIAppBuilder.md) [Next Topic
Accuracy Debugger](https://docs.qualcomm.com/bundle/publicresource/80-62010-1/topics/accuracy_debugger.md)