OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

AI Hub Models

AI Hub Models is a collection of machine-learning models optimized for performance and deployment on Qualcomm devices, accompanied by Python packages and a command-line interface. The tools let developers browse model information, performance and evaluation results, download deployable assets, and use the full package to export, evaluate and run end-to-end demos. Documented runtimes include Qualcomm AI Engine Direct,

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

AI Hub Models — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you need to explore machine-learning models with Qualcomm-device deployment in mind, compare the published performance or evaluation data, or work through a Python-based export and evaluation path.

Who it is for

  • Developers selecting models for applications targeting Qualcomm devices.
  • Python developers who need model details, performance results or downloadable deployment assets.
  • Teams preparing models for compilation, profiling or evaluation on hosted Qualcomm hardware.

Documented uses

  • Browse models and inspect model-specific information, numerics and performance.
  • Download deployable model assets for a selected runtime and precision.
  • Export and evaluate a model through AI Hub Workbench for a hosted Qualcomm device.
  • Run documented end-to-end examples locally with PyTorch or on a hosted Qualcomm device.
Before you choose

Requirements & limits

The repository lists the project under the BSD-3-Clause license. The documented hosted-device export and evaluation workflow requires a Qualcomm ID, AI Hub Workbench access and an API token.

  • Choose the CLI-only package for browsing and asset downloads; the full package includes export, evaluation and demos.
  • The local demo path uses standard PyTorch CPU or GPU kernels and does not provide NPU support; device-accelerated demos use a cloud-hosted Qualcomm device.
  • Runtime operating-system support varies: the README lists Qualcomm AI Engine Direct and ONNX for Android, Linux and Windows, and LiteRT for Android and Linux.
  • Not all models have a cloud-hosted demo; some are documented as PyTorch-only.
  • Python applications emphasize clarity rather than minimum prediction time.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Browse model details and download an asset

Inspect a model’s catalog information and fetch a deployable asset in a chosen runtime and precision.

Before you start

  • Python environment with the qai-hub-models-cli package installed.
  • A model identifier and the desired runtime and precision.

Steps

  1. Install the lightweight CLI package with the command below.

    Documented command
    pip install qai-hub-models-cli

  2. Run qai-hub-models models to browse the catalog.

  3. Use qai-hub-models info mobilenet_v2, numerics mobilenet_v2 or perf mobilenet_v2 to inspect the documented details and results for that example model.

  4. Fetch an asset with the documented example qai-hub-models fetch mobilenet_v2 --runtime tflite --precision float, substituting the intended model and supported options.

Expected outcome

The CLI can display model information, evaluation and performance results, and download a deployable model asset.

Guide limitations

  • Runtime and precision support vary by model and target; check the model’s support details.
  • This guide covers catalog access and asset download, not compiling or running the model.
Documentation 1 ↗
Source-based practical guide

Export and evaluate a model on a hosted Qualcomm device

Configure Workbench access, install model dependencies, and submit a model for device-targeted export and evaluation.

Before you start

  • A Qualcomm ID and Qualcomm AI Hub Workbench access.
  • An API token configured with the qai-hub configure command.
  • The full qai_hub_models package and model-specific dependencies.
  • A model, target runtime, precision and hosted device supported by the project.

Steps

  1. Create a Qualcomm ID, sign in to AI Hub Workbench, obtain an API token, and configure it with qai-hub configure --api_token API_TOKEN.

  2. Install the selected model’s dependencies, following the documented example qai-hub-models install yolov7.

  3. Run the model’s export command with the chosen target runtime, precision and device; the README gives a YOLOv7 example targeting LiteRT on a Samsung Galaxy S25 family device.

  4. Use the model ID printed by export with the corresponding qai-hub-models evaluate command.

Expected outcome

The workflow submits a model for compilation and profiling on a hosted Qualcomm device, then evaluates it against a dataset subset.

Guide limitations

  • The example uses YOLOv7; exact model options and dependencies are model-specific.
  • The guide requires Workbench credentials and a hosted device target. It does not establish execution results for a reader’s particular model or device.
  • The README notes that full-package installation on Windows for Snapdragon X Elite and X2 Elite requires AMD64 Python, not Windows ARM64 Python.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
13 October 2026
Original website title
qualcomm/ai-hub-models · Official homepage title · retained in its source language
How we review sourcesSuggest a correction