1、VP,Product Management,AI/Gen AIQualcomm Technologies,Inc.Dr.VineshSukumarQualcommAI EngineQualcommAIEngineQualcomm AI EngineUp to37%fasterUp to16%better performance per watt*Compared with previous generation.Based on ML Perf Geomean 25.*Based on Ludashi AImark v4 Resnet50 model with internal power m
2、easurement and compared with the previous generation of Snapdragon 8 Elite.LVM and classicAI model accelerationLLM and classicAI acceleration12-core scalar acceleratorLong context support8-core vector accelerator Higher throughput for all coresNew Precision support:INT2&FP8220tokens per second3 bill
3、ion parameter small language model.Peak performance,using Qualcomm advanced parallel decoding technique for select use case with 2K context length.Up toAgentic AI“Hey,can you make a reservation at my moms favorite restaurant around 6 PM?Its her birthday today.Id also like to pick up some flowers and
4、 a cake on the way.Let me know what time I should leave the office and send the directions to my car.”UI and app controlDownload for free on Google Play StoreOptimized forPaage.AI80 TOPSUp to78%faster than previous genWorlds fastest NPUfor laptopsNPU performance is based on UL Procyon AI Computer Vi
5、sion score on Windows 11 OS,run in September 2025.Snapdragon X2 Elite Extreme(X2E-96-100)was tested using a Qualcomm reference design.Snapdragon X Elite(X1E-84-100)was tested using a Qualcomm reference design.Based on manufacturer listed NPU INT8 TOPS specification55AMD Ryzen Al 9 HX 37550AMD Ryzen
6、Al 7 35050AMD Ryzen AI 9 Max+3951348Intel Core Ultra 9 288V13Intel Core Ultra 9 285HXIntel Core Ultra 9 285H38AppleM4 Pro38AppleM438AppleM4 Max80Worlds fastest NPU for laptopsSnapdragonX2 Elite4151719174218662121Snapdragon X2 Elite ExtremeIntel Core Ultra 9 285HAMD Ryzen AI 9 HX 370Intel Core Ultra