Skip to main content
This is the prepare-your-model step of the Choose your journey flow. Before a model runs efficiently on Qualcomm hardware it is converted to an executable format and, for the Hexagon NPU (HTP), quantized to a supported precision. Qualcomm provides several tools to download a preoptimized model, convert and compile your own, fine-tune with custom data, or recover accuracy lost to quantization. Select the tool that matches your starting point.

Use AI Hub to optimize an AI model

Download a preoptimized model, or bring your own and have it compiled, converted, and quantized in the cloud with Qualcomm AI Hub.

Use Qualcomm AI Runtime SDK to optimize an AI model

Convert, quantize, and compile your own models locally with the Qualcomm AI Runtime (QAIRT) SDK.

Recover accuracy with Qualcomm AI Model Efficiency Toolkit (AIMET)

Recover accuracy lost during quantization using post-training quantization (PTQ) and quantization-aware training (QAT).

Fine-tune an AI model with custom data using Edge Impulse

Build, train, or fine-tune models from your own audio, image, and sensor data with Edge Impulse.
Already have a ready-to-run model? You can download a preoptimized LiteRT model from AI Hub and skip straight to Run inference.