INT4 AI Model Transformation on Snapdragon 8 Gen 2

The Snapdragon 8 Gen 2 mobile platform, announced at Snapdragon Summit 2022, is the industry’s first to commercialize INT4 precision.

In the past, Qualcomm successfully enabled transformation of several AI models from floating point 32 (FP32) to INT16 all the way to INT8 without compromising accuracy. Additionally, these improvements gave the artificial intelligence models running on Snapdragon 8 gen 2 the added advantage of higher performance at lower memory consumption.

Now Snapdragon 8 Gen 2 pushes the boundaries, implementing INT4 for higher power savings without compromising accuracy or performance. The exact same model, using INT4 on Snapdragon 8 Gen 2, can be natively processed on our Hexagon Processor much faster than a FP32 model. Alternatively, with the same amount of time, Snapdragon 8 gen 2 can process many more neural network models , which means only more exciting AI experiences at a fraction of the power consumption, so you can do more with one charge.