> ## Documentation Index > Fetch the complete documentation index at: https://qualcomm-3.mintlify.site/llms.txt > Use this file to discover all available pages before exploring further. # AI Models > Browse, download, profile, and manage AI models for your Qualcomm® device — directly from the IDE. ## Overview The **AI Models** section lets you browse the Qualcomm AI Hub model catalog, download models to your host, flash them to your active device, and profile them — all without leaving VS Code. The AI Models section has two tabs: * **MY MODELS** — models you have already downloaded to the host * **EXPLORE MODELS** — the full Qualcomm AI Hub model catalog The active device is shown in the top-right corner of the AI Models page. *** ## My Models When no models have been downloaded yet, **MY MODELS** shows an empty state with an **Explore Models** button: AI Models — My Models empty state Once models are downloaded, each row shows the model name, tags (task, precision, runtime, supported platforms), local file path, and **Profile** / **Flash** action buttons. AI Models — downloaded model with file path and action buttons *** ## Explore Models The **EXPLORE MODELS** tab displays the AI Hub model catalog as a card grid grouped by category (e.g., Computer Vision, Generative AI, Multimodal). Each category heading shows the total number of available models and a **View All** link. AI Models — Explore Models tab with category grid Each card shows: * A preview image with a task badge (e.g., Image Classification, Text Generation) * Model name and short description * **Download** and **View Details** action buttons ### Filtering the Catalog Use the **Device** and **Task** filter dropdowns above the grid to narrow results. Clicking **Device** opens a modal with a search field and device chip selectors (e.g., RB3 Gen2 Vision Kit, IQ-9075 Evaluation Kit, IQ-8275 Evaluation Kit Pro). Select one or more devices, then click **Apply Filter**. AI Models — Device filter modal Clicking **Task** opens a similar modal with task chip selectors (e.g., Depth Estimation, Image Classification, Image Editing, Object Detection, Pose Estimation, Semantic Segmentation, Speech Recognition, Super Resolution). Select one or more tasks, then click **Apply Filter**. AI Models — Task filter modal *** ## Model Detail Page Click **View Details** on any catalog card to open the model detail page. You can also reach this page from **MY MODELS** by clicking the model name. The breadcrumb shows the navigation path: * From **Explore Models**: `Explore Models > > ` * From **My Models**: `My Models > ` The detail page reached from **Explore Models** shows the full benchmark table with columns for each precision/runtime combination across supported devices: AI Models — Explore Models detail page with full benchmark table The detail page reached from **MY MODELS** shows Benchmark Results alongside profiling results columns: AI Models — My Models detail page with benchmark and profiling results The detail page includes: * Model name, task badge, and description * Version, **Model Repository** link, and **Research Paper** link * **Download** button * **Results** table with a **Device** selector * **Technical Details** — model checkpoint, input resolution, parameter count, model size * **Supported Devices** — list of compatible dev kits * **Applicable Scenarios** — suggested use cases * **Licenses** — source model license *** ## Downloading a Model ### Starting a Download Click **Download** on a catalog card or on the model detail page. In the **Download Model** dialog, select the desired **Precision**, **Float**, and **Runtime** values, then click **Start Download**. AI Models — Download Model dialog ### Download Progress While downloading, the model appears in **MY MODELS** with a **Downloading…** label and a progress bar: AI Models — download in progress When the download completes, a toast notification confirms: *"\ downloaded successfully."* The model row updates to show its local file path and the **Profile** and **Flash** action buttons. *** ## Profiling a Model Each model row in **MY MODELS** has a **Profile** split-button. Click the dropdown arrow (▾) to choose from four options: AI Models — Profile split-button dropdown | Option | Description | | ---------------------------- | ------------------------------------------------------------ | | **Profile in AI Hub** | Run the model on the cloud to analyze performance | | **Profile on active device** | Run the model on the connected device to analyze performance | | **View profiling results** | Open the profiling results page | ### Profile on Active Device Selecting **Profile on active device** starts local profiling. AI Models — pushing benchmark model to device AI Models — running measurements on device When complete, a toast notification confirms: *"AOT-GAN profiled locally on IQ-8075 Evaluation Kit."* AI Models — local profiling complete An active device must be set before using **Profile on active device**. See [Devices](../development/devices) for setup instructions. ### Profile in AI Hub #### AI Hub Token The first time you profile a model in AI Hub, the extension prompts you to enter your AI Hub token: AI Models — Enter AI Hub Token prompt Paste your token from [aihub.qualcomm.com/account](https://aihub.qualcomm.com/account) and press **Enter** to confirm or **Escape** to cancel. Selecting **Profile in AI Hub** opens a side panel to configure the cloud profile job. Select the **Device Type** from the dropdown, then click **Start Profile**. AI Models — Profile in AI Hub side panel A notification panel confirms the job was created: *"Profile job created on AI Hub. Results may take a while"* and *"Starting profiling on AI Hub."* AI Models — AI Hub profile job started notifications ### Viewing Profiling Results Select **View profiling results** to open the results page from the model detail view. The breadcrumb shows `My Models > `. AI Models — profiling results page with all three result columns The results table header shows the Component, Precision, Runtime, and Device used for profiling. Columns show results from all available sources: | Column | Description | | ----------------------------------- | ------------------------------------------------ | | **Benchmark Results** | Published AI Hub reference values for the model | | **Profile Results – AI Hub** | Values from the cloud AI Hub profiling job | | **Profile Results – Active Device** | Values measured on your locally connected device | Each value in the **Profile Results** columns has a colored arrow indicating performance relative to the benchmark: * **Red** — worse than benchmark * **Green** — better than benchmark A **More Details** link in each column opens the full AI Hub profiling report. *** ## Flashing a Model to the Device Click **Flash** on a model row in **MY MODELS** to push the model to the active device. AI Models — transferring model files to device AI Models — pushing model to device with progress bar When complete, a toast confirms: *"AOT-GAN model pushed to IQ-8075 Evaluation Kit successfully."* AI Models — model pushed successfully with Pushed badge An active device must be set before flashing. The active device is shown in the top-right corner of the AI Models page. See [Devices](../development/devices) for setup instructions. *** ## Removing a Model Click the **⋮** overflow menu on the right side of a model row in **MY MODELS** and select **Remove** to delete the model and all its profiling results from the host. AI Models — Remove option in overflow menu **Remove** deletes the local model file and all associated profiling results. The model can be re-downloaded from the **Explore Models** tab.