> ## Documentation Index
> Fetch the complete documentation index at: https://dragonwingdocs-staging.qualcomm.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Run on-device Generative AI (GenAI)

> Prepare, run, and develop generative AI applications on Qualcomm Dragonwing IoT platforms using Genie, Qualcomm AI Hub, and the QAIRT SDK.

Qualcomm® Dragonwing™ products bring generative AI (GenAI) to edge devices without relying solely on cloud infrastructure, lowering latency, enhancing privacy, and reducing costs for diverse IoT applications.

<Note>
  'GenAI use-cases' are not supported on Dragonwing™ `IQ-2390` and Dragonwing™ `IQ-615`.
</Note>

The following image illustrates the GenAI architecture, including supported use cases and applications, GenAI frameworks, and the lower-level backends and libraries.

<img src="https://mintcdn.com/qualcomm-staging/blZpKs_b9On5nj-y/archives/2.1/Key-Documents/AI-Developer-Workflow/_images/genai-arch.png?fit=max&auto=format&n=blZpKs_b9On5nj-y&q=85&s=c35724ccbbef801c24f1d812a7d43f3f" alt="GenAI architecture diagram" style={{ width: "150%", maxWidth: "none", marginLeft: "-25%" }} width="2783" height="1606" data-path="archives/2.1/Key-Documents/AI-Developer-Workflow/_images/genai-arch.png" />

Dragonwing IQ products integrate a CPU, GPU, and Hexagon NPU for heterogeneous computing, optimized for large language models (LLMs), vision, and multimodal AI tasks. Leading GenAI models — including Llama, Whisper, Stable Diffusion, and LLaVA — are supported for IoT scenarios such as robotics, retail, and industrial automation.

For responsive LLM performance, Dragonwing IQ supports multi-billion parameter models with techniques such as prefill acceleration and self-speculative decoding. To enable bring-your-own-model (BYOM) workflows and streamline deployment, use Qualcomm AI Hub, Qualcomm AI Runtime SDK execution providers, and Qualcomm Generative AI Inference Extensions (Genie).

## Why on-device GenAI matters

Generative AI on Dragonwing delivers context-aware, adaptive, and privacy-preserving solutions at the edge, enabling automation, personalization, and operational efficiency for use cases such as autonomous drones, predictive maintenance, multimodal agents, and on-device retrieval-augmented generation (RAG).

## Get started

Select the option that matches your GenAI workflow.

<CardGroup cols={2}>
  <Card title="Prepare a GenAI model" icon="wand-magic-sparkles" href="../topic/prepare-genai-model">
    Prepare a GenAI model for execution on Qualcomm Dragonwing IoT devices.
  </Card>

  <Card title="Run a GenAI model" icon="robot" href="../topic/run-genai-model">
    Run a prepared GenAI model on a Qualcomm Dragonwing IoT device.
  </Card>

  <Card title="Use GenAI models with Genie" icon="brain" href="../topic/use-genai-model-with-genie">
    Prepare, manage, and execute GenAI models with Qualcomm Generative AI Inference Extensions (Genie).
  </Card>

  <Card title="Develop a GenAI application" icon="rocket" href="../topic/develop-genai-app">
    Build your own GenAI application based on a Qualcomm sample application.
  </Card>
</CardGroup>
