Zaid Ahmed

Tom Gunter, Zirui Wang, Chong Wang et al.

We present foundation language models developed to power Apple Intelligence features, including a ~3 billion parameter model designed to run efficiently on devices and a large server-based language model designed for Private Cloud Compute. These models are designed to perform a wide range of tasks efficiently, accurately, and responsibly. This report describes the model architecture, the data used to train the model, the training process, how the models are optimized for inference, and the evaluation results. We highlight our focus on Responsible AI and how the principles are applied throughout the model development.

Zaid Ahmed, Omar A. Khan, Hyeongil Nam et al.

Extended Reality (XR) enables immersive capture and re-experience of personal memories, yet how interface representations shape these experiences remains underexplored. We examine how users relive and share XR memories through three interaction approaches: (1) physical memory-linked objects, (2) virtual memory-linked objects, and (3) a conventional virtual gallery interface. In a within-subjects study (N=24, 12 pairs), participants captured shared experiences using 360° video and later accessed and shared these memories across the three interfaces. We analyzed open-ended qualitative responses focusing on perceived value, enjoyment, usability, emotional attachment, and social connection. The findings reveal trade-offs: physical objects fostered stronger social connection and conversation through tangible exchange; virtual objects balanced engagement and usability; and the gallery interface was efficient but less personal. These results suggest that object-based representations, physical and virtual, support key social dimensions of XR memory experiences, offering lessons for designing future systems that emphasize shared meaning and interpersonal connection.

Ethan Li, Anders Boesen Lindbo Larsen, Chen Zhang et al. · apple-ml, cmu

We introduce two multilingual, multimodal foundation language models that power Apple Intelligence features across Apple devices and services: i a 3B-parameter on-device model optimized for Apple silicon through architectural innovations such as KV-cache sharing and 2-bit quantization-aware training; and ii a scalable server model built on a novel Parallel-Track Mixture-of-Experts PT-MoE transformer that combines track parallelism, mixture-of-experts sparse computation, and interleaved global-local attention to deliver high quality with competitive cost on Apple's Private Cloud Compute platform. Both models are trained on large-scale multilingual and multimodal datasets sourced via responsible web crawling, licensed corpora, and high-quality synthetic data, then further refined with supervised fine-tuning and reinforcement learning on a new asynchronous platform. The resulting models support several additional languages while understanding images and executing tool calls. In public benchmarks and human evaluations, both the server model and the on-device model match or surpass comparably sized open baselines. A new Swift-centric Foundation Models framework exposes guided generation, constrained tool calling, and LoRA adapter fine-tuning, allowing developers to integrate these capabilities with a few lines of code. The latest advancements in Apple Intelligence models are grounded in our Responsible AI approach with safeguards like content filtering and locale-specific evaluation, as well as our commitment to protecting our users' privacy with innovations like Private Cloud Compute.