# qualcomm/ai-hub-models

Qualcomm® AI Hub Models is our collection of state-of-the-art machine learning models optimized for performance (latency, memory etc.) and ready to deploy on Qualcomm® devices.

Repository: https://github.com/qualcomm/ai-hub-models
Canonical: https://ross.abutalabs.com/products/ai-hub-models
Homepage: https://aihub.qualcomm.com/models
Language: Python
License: BSD-3-Clause
License Family: permissive
Topics: deeplearning, demos, inference, inference-api, inference-engine, machine-learning, machinelearning, onnx, pytorch, qnn, tensorflow-lite
Last push: 2026-09-03T01:46:38+00:00

## Health v2 (maintenance only)
Score: 89/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 100, release rhythm 87, longevity 70
- inputs: {"age_days": 987, "days_push": 0, "days_rel": 8, "gap_med": 12, "n_releases_24m": 64}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1195, forks 208 (observed 2026-09-03T02:15:17.194493+00:00)

## What it is
Qualcomm AI Hub Models is a curated collection of 300+ state-of-the-art machine learning models (vision, audio, speech, generative AI) pre-optimized for deployment on Qualcomm Snapdragon devices. It ships as a Python package and CLI for browsing, exporting, compiling, and profiling models to runtimes like TensorFlow Lite, ONNX, and Qualcomm AI Runtime (QNN).

## Use cases
- deploy a machine learning model on a Snapdragon phone
- convert pytorch model to tflite for mobile
- benchmark model latency on qualcomm NPU
- download pre-optimized yolov7 for on-device inference
- run image classification on android device
- profile model performance on hexagon NPU
- find models optimized for qualcomm chipsets

## When to choose
- You are deploying ML models to Qualcomm Snapdragon / Dragonwing hardware and want pre-validated, hardware-optimized assets
- You need on-device performance metrics (latency, memory, compute unit utilization) for Qualcomm chipsets
- You want a CLI/Python API to export models to TFLite, ONNX, or QNN runtimes

## When to avoid
- You target non-Qualcomm hardware such as Apple Silicon, NVIDIA GPUs, or generic x86 servers
- You need to train or fine-tune models rather than deploy them
- You require a fully offline workflow - many features (compilation, profiling) require Qualcomm AI Hub Workbench cloud access

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, llm-inference, image-processing, speech-recognition, computer-vision, cli, sdk
- domain: machine-learning, deep-learning, artificial-intelligence, mobile-development, computer-vision, embedded-systems, developer-tools
- platform: python, windows, cli, embedded
- tags: qualcomm, on-device-ai, model-optimization, tflite, onnx, qnn, snapdragon, npu, model-deployment, pytorch, android, linux, mobile

## Member repositories
- qualcomm/ai-hub-models (main) score 89

## Provenance
- Observed fields: from GitHub, fetched 2026-09-03T02:15:17.194493+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:22:43.127682+00:00, confidence not recorded.
  - readme: https://github.com/qualcomm/ai-hub-models (fetched 2026-09-03T02:15:17.194493+00:00, sha 11457d0c9067)
  - homepage: https://aihub.qualcomm.com/models (fetched 2026-08-29T12:30:23.494929+00:00, sha 78fed1e9beac)
  - site_page: https://workbench.aihub.qualcomm.com/docs (fetched 2026-08-29T12:30:23.507473+00:00, sha 3fa960f80ada)
  - site_page: https://aihub.qualcomm.com/geniex (fetched 2026-08-29T12:30:23.503994+00:00, sha b3917f877526)
  - site_page: https://geniex.aihub.qualcomm.com/ (fetched 2026-08-29T12:30:23.505790+00:00, sha 96e548aca624)
  - site_page: https://workbench.aihub.qualcomm.com (fetched 2026-08-29T12:30:23.509304+00:00, sha 0c500b72bc50)
- Data as of 2026-08-30T08:39:29.467469+00:00.