During the Mobile World Congress in 2024, Qualcomm revealed more than seventy-five new large language models that are tailored for the Snapdragon platform. Additionally, the company has confirmed that some of these models will power generative AI features in software-defined cars, PCs, IoT, XR devices, and next-generation smartphones.
“The Qualcomm AI Hub provides developers with a comprehensive AI model library to quickly and easily integrate pre-optimised AI models into their applications, leading to faster, more reliable, and private user experiences,” said Durga Malladi, senior vice president and general manager of technology planning and edge solutions at Qualcomm Technologies, Inc.
A collection of “pre-optimized AI models” can be accessed through the new Qualcomm AI Hub and used on Qualcomm and Snapdragon-powered devices. At the moment, it contains more than 75 models, some of which are well-known generative AI models like Baichuan 7B, ControlNet, Whisper, and Stable Diffusion. These models are claimed to provide on-device generative AI experiences with reduced power consumption and memory utilization with further optimization. Additionally, they have been optimized with up to 4 times faster inferencing time for the Qualcomm AI Engine (NPU, CPU, and GPU). These models are available to developers through the Qualcomm AI Hub as well as other platforms like Hugging Face and GitHub.
Large Language and Vision Assistant (LLaVA), a customized model with over 7 billion parameters that can run natively on an Android smartphone, was demonstrated by the company at the event. In a similar vein, the business unveiled the first 7+ billion parameter model in the world that can handle audio and text input.
Additionally, Qualcomm is showcasing some of the AI features found on smartphones running the Snapdragon 8 Gen 3 processor, including the HONOR Magic6 Pro (which creates videos and calendars using AI), the OPPO X7 Ultra (which erases objects from images), and the Xiaomi 14 Pro (which expands images using AI).
