Qualcomm, at the Mobile World Congress 2024, has announced over 75 new large language models optimised for the Snapdragon platform. The company has also confirmed that some of these models will be powering generative AI capabilities across next-generation smartphones, PCs, IoT, XR devices, and software-defined vehicles. “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. The new Qualcomm AI Hub is a library of "pre-optimised AI models" that can be deployed on devices powered by Qualcomm and Snapdragon platforms. It currently has over 75 different models, including some popular generative AI models like Whisper, ControlNet, Stable Diffusion, and Baichuan 7B. With additional optimisation, these models are said to deliver on-device generative AI experiences while consuming less power with lower memory utilisation. They have also been optimised for Qualcomm AI Engine (NPU, CPU, and GPU) with up to 4x faster inferencing time. Developers can access these models via the Qualcomm AI Hub or from other platforms like GitHub and Hugging Face. At the venue, the company demonstrated Large Language and Vision Assistant (LLaVA), a custom model with over 7 billion parameters, which can natively run on an Android smartphone. Similarly, the company also showcased the world's first 7+ billion parameter model that can accept both text and audio input. Qualcomm is also demonstrating some of the AI features available on Snapdragon 8 Gen 3-powered smartphones such as HONOR Magic6 Pro (AI-powered video creation and AI-powered calendar creation), OPPO X7 Ultra (image object eraser), and Xiaomi 14 Pro (AI-generated image expansion).
