Apple Unveils Third-Generation Foundation Models with Enhanced AI Capabilities Apple has introduced its third generation of foundation models, marking a significant leap in its efforts to integrate advanced artificial intelligence into its ecosystem. These models, developed in collaboration with Google and NVIDIA, are designed to enhance user experiences through improved capabilities, privacy-focused architecture, and optimized performance across both on-device and server-based systems. The new foundation models, collectively referred to as Apple Foundation Models (AFM), include five distinct variants tailored for different applications, from voice and text processing to image generation and complex reasoning tasks. The third-generation models feature two on-device variants: AFM 3 Core and AFM 3 Core Advanced. AFM 3 Core represents an upgrade to Apple’s existing 3-billion-parameter dense model, offering enhanced quality and efficiency. AFM 3 Core Advanced, however, stands out as the most powerful on-device model, leveraging a novel sparsely activated architecture. This model is natively multimodal, enabling features like expressive voices and high-accuracy dictation. Built on cutting-edge research, AFM 3 Core Advanced uses a sparse architecture that activates only 1 to 4 billion parameters at a time, depending on the request. This design allows the model to operate efficiently on Apple’s most capable silicon systems while maintaining performance. Complementing the on-device models are three server-based variants running on Apple’s Private Cloud Compute infrastructure.#apple #google #nvidia #private_cloud_compute #apple_foundation_models
