Tag
A cookbook repository by Liquid4All with examples, end-to-end tutorials, and applications for building with Liquid AI's open-weight LFMs and the LEAP SDK on laptops, mobile, and edge devices.
An 8 billion parameter AI model can now run offline on an iPhone 15 through a dedicated app, enabling on-device inference without internet.
AiOS San Francisco is an in-person meetup for AI and Apple platform enthusiasts, featuring talks and demos from Wendy Labs and MiniMax on Swift-powered AI, robots, embedded devices, and the MiniMax H3 video-generation model running locally on a Mac.
The paper compares five parameter-efficient fine-tuning methods on four small language models for on-device personalization, finding LoRA+ best for energy efficiency and QLoRA best for memory-limited deployment.
LiveTranscriber is an open-source iOS app that runs Whisper, Qwen3-ASR, Nemotron, MOSS, and Qwen3 fully offline on iPhone, offering speech transcription, multi-speaker support, summaries, and real-time translation. The developer shares the engineering challenges and invites feedback from ASR and on-device AI communities.
Liquid AI announces a partnership with MacPaw to bring on-device AI to Mac users, designing specialized Liquid Foundation Models for macOS AI assistance.
ProcAgent is a fully on-device, agentic, vision-based procedural assistant that uses a propose-and-verify architecture for real-time adaptive guidance on an NVIDIA Jetson AGX Orin. It supports both reactive and proactive modes with human-in-the-loop confirmation, achieving responsive interaction and positive user study ratings.
Microsoft announces general availability of Video Super Resolution for NPU and CPU, using on-device AI to enhance lower-resolution video quality.
Google AI Edge Gallery is an experimental beta app that lets developers test and evaluate generative AI models on-device without building a full mobile app. It supports model downloads, custom loading, parameter tuning, hardware benchmarks, and multimodal workflows, and is open-source for Android, iOS, and macOS.
Google Cloud deployed a Coaching Agent natively inside the new FIA Formula E GEN4 car, using extreme edge compute with a Raspberry Pi 5 and Google Pixel 10 Pro XL to process microsecond-accurate telemetry entirely on-device, without remote data centers.
Google announced Gemma 4 E2B optimized for the Pixel 10's TPU, enabling on-device multimodal AI capabilities like offline chat, image recognition, and audio transcription.
Prism ML is in talks with Apple to deploy model-shrinking technology that would allow powerful AI models to run directly on iPhones, improving privacy and reducing reliance on cloud inference.
LokalBot v0.2.0 update adds local Agent Mode, MCP server for meeting queries, system-wide dictation, and improved resource management, maintaining 100% on-device and open-source status.
Prediction that within two years, local AI models will be widely accessible, with Apple's Mac Studios supporting large RAM for on-device AI; encourages preparing now by using local AI tools.
This literature review introduces MELT, a benchmark for evaluating large language models (LLMs) on mobile devices, highlighting the challenges and opportunities of running transformers on phones.
Apple Silicon exec Doug Brooks explains the surging demand for Mac mini and Mac Studio for AI agent workloads, highlights Apple's chip design philosophy integrating neural engines, and discusses the shift toward on-device AI for privacy and cost reasons while envisioning a hybrid future.
The PyTorch Foundation supported the ExecuTorch Hackathon in San Francisco, where over 100 participants built real-time on-device AI applications using PyTorch and ExecuTorch on Snapdragon-powered Samsung Galaxy S25 Ultra devices. Winning projects included SafeScreen AI, SixthSense, and Toddle AI, showcasing local execution benefits for responsiveness, privacy, and offline capability.
A tweet by Ahmad Osman expressing a goal to make local AI the default.
Google Research introduces a new architecture using frozen Multi-Token Prediction to accelerate Gemini Nano models on Pixel devices, significantly improving speed and energy efficiency for on-device AI features.
Apple plans to skip M6 Pro/Max chip variants to accelerate development of M7 chips for enhanced on-device AI, aiming for a 2027 release. The move responds to competition from Nvidia, AMD, Intel, and Qualcomm in AI processing.