Qualcomm launches two new smartphone chips with emphasis on AI
Qualcomm unveiled two flagship smartphone processors at its Snapdragon Summit: the Snapdragon 8 Elite Gen 6 and the higher-end Snapdragon 8 Elite Extreme Gen 6, both built to accelerate on-device AI. The company says the chips include new sensing hubs for running models up to 200 million parameters and that the Extreme variant can run a 30-billion-parameter mixture-of-experts model locally.
Why It Matters
These chips push more advanced AI workloads and privacy-sensitive personal agents onto phones rather than cloud servers by increasing on-device model capability and efficiency. That shift could influence how smartphone makers design features and services that rely on local AI inference.
Key Facts
- Chips announced: Snapdragon 8 Elite Gen 6 and Snapdragon 8 Elite Extreme Gen 6
- Sensing hub model size: Up to 200 million parameters
- Extreme model local capability: Can run a 30-billion-parameter mixture-of-experts (MoE) model locally
- Apple comparison: Apple released a 20-billion-parameter MoE model at WWDC in June
- Camera/video capabilities (Extreme): Supports 8K at 60fps and 4K at 240fps; enables Advanced Professional Video (APV) codec
Qualcomm introduced two new flagship mobile processors, the Snapdragon 8 Elite Gen 6 and the Snapdragon 8 Elite Extreme Gen 6, positioning both around expanded on-device AI capabilities. Each chip includes a redesigned sensing hub Qualcomm says can execute compact models of up to 200 million parameters, enabling features such as local speaker differentiation, personal transcription or "scribe" functions, and usage-based memory for more personalized suggestions and task automation. The company also said the new platform supports running a complete voice-in and voice-out agent locally.
The Snapdragon 8 Elite Gen 6 features a new accelerator intended to increase inference efficiency for models on the device. Qualcomm highlighted that the higher-end Snapdragon 8 Elite Extreme Gen 6 can run a 30-billion-parameter mixture-of-experts (MoE) model locally; because MoE architectures activate only a subset of parameters per task, the company frames this as a way to handle very large models without requiring the entire set of parameters to be active for each inference. For context, Qualcomm contrasted the capability with Apple’s announcement in June of a 20-billion-parameter MoE model.
Beyond AI, Qualcomm said the new CPUs provide more granular pixel-level camera control to support pro-oriented imaging workflows, along with improved stabilization and motion understanding. The Extreme variant specifically supports 8K video recording at 60 frames per second and 4K at 240fps for high-frame-rate capture, and it brings support for Qualcomm’s Advanced Professional Video (APV) codec. Both chips also include AI-driven audio features such as vocal enhancement and noise reduction, plus a "voice bubble" technology designed to isolate a user’s voice during calls.
At the event, Motorola revealed the Motorola Signature 27 as a device powered by the Snapdragon 8 Elite Extreme Gen 6, with general availability planned for later this year. Qualcomm said it is working with more than 40 AI device projects as part of the broader push to put advanced AI features onto mobile handsets.
Keep Reading
Waymo’s latest expansion strategy: teenagers
OpenAI launches GPT-6 Sol and Luna, boasting lower cost and fewer mistakes
a16z is challenging Silicon Valley’s love for drop-outs by launching a school
Hacking group ShinyHunters claims it breached the FBI, stole agents’ and applicants’ data
Original source: TechCrunch