Apple Core ML
Apple's framework for running ML models on-device — hardware-accelerated via Neural Engine on iPhone, iPad, and Mac.
Visit Apple Core ML →AI that runs on the machine in front of you
On-device inference is the part of AI that does not send your data anywhere. This is a directory of the runtimes, the neural accelerators inside consumer hardware, and the platform layers that make a phone or laptop run a model locally.
8 listings · 1 categories · updated September 2026Generated from our master catalogue on the build date shown, listed alphabetically. Not a ranking, not paid placement, nothing solicited. Every outbound link is checked before publication and entries our checker records as dead are removed automatically. Corrections: [email protected]
Advertisement
Browse on-device AI
Every entry is generated from our master catalogue and its link verified before this page was built. Alphabetical; nothing here is a ranking.
Apple Intelligence
Apple's on-device AI system for iPhone, iPad, and Mac — Private Compute Cloud for privacy-preserving AI features.
Visit Apple Intelligence →Arm ML
Arm's ML platform and Ethos NPU IP — the AI compute backbone of most mobile and IoT chips.
Visit Arm ML →Axelera AI
European edge AI chip startup — Metis AI Platform using phase-change memory for ultra-efficient inference.
Visit Axelera AI →Google Gemini Nano
Google's smallest Gemini model for on-device inference — runs locally on Android devices for offline AI features.
Visit Google Gemini Nano →Kneron
Edge AI chip and solution provider — KL series NPUs for smart cameras, IoT devices, and embedded systems.
Visit Kneron →Microsoft Copilot+
Microsoft's Copilot+ PC initiative — NPU-accelerated AI features built into Windows for next-gen hardware.
Visit Microsoft Copilot+ →TensorFlow Lite
Google's lightweight inference framework for mobile and embedded — cross-platform model deployment for on-device AI.
Visit TensorFlow Lite →No matches. Try a different search.
Advertisement
Questions
Short, direct answers.
What does on-device AI actually mean?
The model runs on your phone, laptop or an embedded device instead of on a server. Nothing about the request leaves the hardware, which removes both the round trip and the privacy question at the same time.
What is an NPU and do I need one?
A neural processing unit is silicon dedicated to the maths models need, sitting alongside the CPU and GPU. It matters most for battery life: an NPU does the same inference at a fraction of the power, which is why it appears in phones and laptops rather than in servers.
Why is on-device slower than a cloud API?
Consumer hardware has far less memory bandwidth than a data-centre accelerator, and the models that fit are smaller. Expect capable rather than frontier results. The trade is latency and privacy against raw capability.
Is my data really private on-device?
Only if the application actually keeps it there. Several vendors run a hybrid model that silently escalates harder requests to a server. Check the vendor's own description of when it leaves the device rather than assuming the label covers everything.