Mobile AI | On-Device AI

Qualcomm AI is transforming industries by accelerating global commercialization of AI.

Qualcomm AI makes the impossible possible with an expansive portfolio of platform solutions that gives our partners flexibility and reliability in designing high-quality AI apps and experiences.

The Qualcomm AI Engine is powering the future of industries.

Experience mobile intelligence like never before with our brand-new 6th Gen Qualcomm® Artificial Intelligence (AI) Engine—uniting AI with everything from photography to gaming to connectivity. The completely re-engineered Qualcomm® Hexagon™ 780 Processor features a fused AI-accelerator architecture and brings the total Qualcomm AI Engine performance up to an astonishing 26 TOPS.

The 6th generation AI engine also has a new Tensor Accelerator packing 2 times compute capacity and Scalar Accelerator with 50% performance improvement. Up to 3 times performance per watt improvement. A brand new, 16 times larger shared AI memory with up to 1000 times hands-off-time improvement in certain use cases, and a contextually aware 2nd Gen Qualcomm® Sensing Hub with dedicated AI processor and truly on-demand AI.

Learn more

Qualcomm Hexagon 6th Gen AI Engine

Nov 24, 2020

1:01

AI tools of the trade.

Develop cutting-edge on-device applications and experiences with our Qualcomm AI Engine plus our AI products and hardware tools.

Neural Processing SDK

Save development time and effort with the Qualcomm Neural Processing SDK for artificial intelligence, designed to optimize performance of trained neural networks on Snapdragon-enabled devices.

Snapdragon Developer Tools

Our Snapdragon Developer Tools help optimize application development with tools that identify app performance problems, improve battery life, and more.

Hexagon DSP SDK

Deliver premium interactive user experiences in less time with the cutting-edge, hardware-enabled multimedia feature set in our Hexagon SDP SDK—designed to help overcome typical development hurdles including lack of battery and insufficient processing power.

AI Model Efficiency Toolkit

The Qualcomm AI Model Efficiency Toolkit (AIMET) is open source on GitHub to collaborate with other leading AI researchers and to provide a simple library plugin for AI developers to utilize for breakthrough model efficiency performance.

We’re defining the future of AI.

Artificial Intelligence (AI) is poised to revolutionize our world and transform every aspect of daily life. And Qualcomm Technologies is working to make AI portable and accessible to the masses. We’re striving to improve connectivity and performance and reduce energy and costs. Our process is built on a comprehensive approach to AI research and development that results in sustainable, comprehensive AI solutions.

Learn more

6th Generation Qualcomm AI Engine Deep Dive

Qualcomm Hexagon 780

The brand-new Hexagon 780 processor is completely redesigned and features our biggest leap in architecture and performance in years. We call it the fused AI accelerator architecture. In past generations, we used scalar, vector, and tensor accelerators. For this new generation, we’re removing the physical distances between the accelerators and fusing them together, so it’s now on one big AI accelerator. We’ve also added a dedicated large shared memory across the three different accelerators, so they can share and move data efficiently - up to 1000x faster in certain use cases.

Qualcomm Kryo 680

Provides up to a 25 percent increase in overall CPU performance over Snapdragon 865.

Qualcomm Adreno 680 GPU

Offers a 43 percent AI performance boost and includes new instruction sets like 4-input mixed precision dot product and wave matrix multiply for faster floating-point calculation.

The 2nd Generation Qualcomm Sensing Hub

With Snapdragon 888, we’re introducing the 2nd generation Qualcomm Sensing Hub, and it’s smarter than ever. With a dedicated always-on, low-power AI processor and we’re seeing a mind-blowing 5x AI performance improvement. The extra AI processing power on the Qualcomm Sensing Hub allows us to offload up to 80 percent of the workload that usually goes to the Hexagon processor, so that we can save even more power. All the processing on the Qualcomm Sensing Hub is at less than 1 milliamps (mA) of power consumption.

Featured videos

Qualcomm & Tetras AI Supermovie Demo – Fencing

Jan 27, 2021