MediaTek has introduced the Dimensity 9600 Pro, a processor built on a 2-nanometer process. The chip pairs eight high-performance CPU cores with two neural processing units, aimed at running AI workloads locally on the device rather than in the cloud. MediaTek has not yet detailed which phones or devices will use it.
A revised arXiv preprint examines how small language models deployed on edge devices can be personalized over time without losing prior knowledge. The work focuses on monitoring stability during sequential adaptation, a known risk when models are updated repeatedly. It is a research contribution rather than a product or model release.
A revised arXiv paper proposes adaptive context management to reduce the memory burden of running AI agents locally on devices. The authors note that agent workloads inflate context size through large static tool schemas and long interaction histories, which strains the limited memory of phones and similar hardware. The work aims to make personalized, low-latency on-device assistance more practical under those constraints.
Researchers introduced LEXIC, a small recurrent neural network that estimates whether a reader understood a text by analyzing eye fixation patterns alongside word frequency and word length. The model is designed to run on-device with a compact footprint, which could enable reading interfaces that adapt to a user's comprehension in real time.
Fashion app Daydream has introduced features built on Apple Intelligence following the release of iOS 27. Users can turn photos of outfits saved in their camera roll into shoppable results, and search for products through Siri without opening the app. The update ties Daydream's visual search to Apple's on-device AI capabilities.
Perplexity has brought its Portable Computer agent to Windows, where it runs models locally on the device rather than in the cloud. The tool is a local variant of the company's Perplexity Computer agent, which decomposes and executes multi-step tasks on its own. NVIDIA RTX GPUs provide the acceleration, and keeping processing on the machine means sensitive data does not leave the PC.
Researchers introduce X-CoSD, a distributed inference framework in which a small on-device model drafts tokens and a larger server-side LLM verifies them. The method tackles the mismatch between the two models' vocabularies while cutting the communication cost between device and server. The work aims to make collaborative speculative decoding practical for latency-sensitive edge deployments.
A new arXiv paper examines how schema design affects the ability of small language models to turn a driver's natural-language requests into accurate vehicle function calls. Because these assistants must run on-device, the authors weigh strict memory and latency limits and argue that readable schemas outperform fixed-key formats for this task.
Apple's chief executive John Ternus argued that the iPhone remains the strongest option for AI features, pointing to the company's on-device models. He said running models locally gives users better privacy protection. The remarks position Apple's hardware as its main AI differentiator.
Hugging Face introduced a collection of more than 200 WebGPU kernels intended to speed up AI workloads running locally in the browser or on user devices. The library is published under the huggingface/kernels namespace, giving developers reusable GPU-accelerated operations for client-side inference. It targets the growing interest in running models without sending data to remote servers.
WHY IT MATTERS ↘Reusable, standardized WebGPU kernels lower the engineering cost of client-side inference, making browser and on-device deployment viable for more teams without custom GPU work. That shifts some inference demand away from cloud APIs toward local hardware, weakening vendor lock-in but also complicating model governance since data and weights increasingly live outside the provider's control.