AI News Feed
Market watch
Gravwell 5.10 Adds Five AI Agents That Gather Their Own Investigation ContextPigment launches AI-generated interfaces to broaden business planningAnt Group stresses trusted AI agents as finance and engineering tools launchAWS Elemental Adds AI-Powered Multiview and Ad Tools Ahead of IBC 2026Fable's Style-Weaving Combat Aims to Be Approachable and Faithful, Playground Games SaysQualcomm Teases On-Device AI Agent Push With Next Snapdragon Elite NPU
Gravwell 5.10 Adds Five AI Agents That Gather Their Own Investigation ContextPigment launches AI-generated interfaces to broaden business planningAnt Group stresses trusted AI agents as finance and engineering tools launchAWS Elemental Adds AI-Powered Multiview and Ad Tools Ahead of IBC 2026Fable's Style-Weaving Combat Aims to Be Approachable and Faithful, Playground Games SaysQualcomm Teases On-Device AI Agent Push With Next Snapdragon Elite NPU
AI Chips & Compute

Qualcomm Teases On-Device AI Agent Push With Next Snapdragon Elite NPU

Qualcomm is teasing a next-generation Hexagon NPU for its upcoming Snapdragon Elite chips, claiming it can run on-device AI agents and Mixture-of-Experts models with up to 30 billion parameters.

AI News Feed · 2026-09-10 22:51

Qualcomm is set to reveal the next generation of flagship chips, including two Elite variants of what is likely to be the Snapdragon 8 Gen 6, on September 22 at the next Snapdragon Summit. The final names have not been disclosed. Qualcomm has been releasing details about the platform in stages ahead of the event.

The company previously said the upcoming chips will include the first smartphone CPU to breach the 5GHz mark and an upgraded GPU with improved frame upscaling technology. Its latest teaser focuses on the Hexagon NPU, which Qualcomm claims will bring specialized upgrades to respond to queries much faster.

Qualcomm said two key highlights of the coming NPU are a new Element Accelerator and much larger shared memory. The company is stressing on-device agentic AI workflows, arguing that such hardware can support more advanced AI tasks without relying on cloud processing.