Qualcomm launches Snapdragon 8 Elite Gen 6 chips built for on device AI agents
The Extreme model can run a 30 billion parameter mixture of experts model locally as Qualcomm pushes more AI processing onto smartphones.
Qualcomm has launched the Snapdragon 8 Elite Gen 6 and Snapdragon 8 Elite Extreme Gen 6, its latest flagship smartphone processors with a focus on running more capable AI agents directly on devices.
The chips include an upgraded sensing hub capable of running models with up to 200 million parameters.
Qualcomm said the system can support features including local transcription, speaker identification and personalized memory that learns from how a user interacts with their device.
Qualcomm said the architecture can also support complete voice in and voice out agents locally, allowing more AI interactions to run without relying entirely on cloud infrastructure.
The higher end Snapdragon 8 Elite Extreme Gen 6 can run a 30 billion parameter mixture of experts model on device. The architecture activates only a portion of the model for each task.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Qualcomm said a 30 billion parameter model could activate roughly 3 billion parameters during each token generation step, reducing compute and memory requirements.
The capability puts Qualcomm alongside a broader industry push toward larger local AI models. Apple introduced a 20 billion parameter sparse on device model in June that activates between 1 billion and 4 billion parameters depending on the request.
Qualcomm also added new camera and audio capabilities. The Extreme chip supports recording at up to 8K 60fps and 4K 240fps, while both processors use AI for vocal enhancement and noise reduction.
Motorola said its upcoming Signature 27 smartphone will be among the first devices powered by the Snapdragon 8 Elite Extreme Gen 6.
The launch reflects Qualcomm’s broader bet that smartphones will remain a major platform for personal AI agents even as companies experiment with dedicated AI hardware.