Running a DeepSeek Model

Qualcomm branded products are products of Qualcomm Technologies, Inc. and/or its subsidiaries. Qualcomm Academy is a program of Qualcomm Technologies, Inc. and/or its subsidiaries.

In this workshop, you will learn how to run DeepSeek LLM locally on Snapdragon(R) X Elite using ONNX Runtime and Qualcomm Hexagon NPU for blazing-fast AI performance.

two people coding and learning at a computer

Running DeepSeek Model on Qualcomm Hexagon Processor

In this build-along style workshop, you will gain a step-by-step tutorial covering the following topics:

  • Setting up Windows on Snapdragon Python environment
  • Installing ONNX Runtime with Hexagon execution provider
  • Deploying DeepSeek R1 (7B) model with tokenizer
  • Optimizing inference with burst mode, low power mode, and caching
  • Implementing greedy sampling, temperature, top-k, and repetition penalty
  • Running chatbot in Jupyter Notebook and command-line app
  • Boosting data privacy, security, and efficiency by owning your LLM and running it on-device