Open source large language models, on your data that doesn't leave your device. No cloud, no account, no telemetry.
Every token generated on your device. Works in aeroplane mode.
Cloud AI sends every word you type to somebody else's computer. Pathway Lite takes the opposite position. The model lives on your machine, the inference happens on your silicon, and your conversations are stored in a database only you can open.
Models run natively on Apple Silicon through MLX. After the one-time model download, no internet connection is required. Chat on a plane, in a tunnel, or with the Wi-Fi switched off entirely.
There is no account, no tracking, and no telemetry. Your chats, documents and projects live in a local database on your Mac. We could not read your conversations even if we wanted to, because they never reach us.
Pathway Lite is sandboxed and notarised, distributed through the Mac App Store. Every model download is integrity-checked against SHA-256 hashes before it is ever loaded, and the app's web layer is blocked from the network by policy.
Pathway Lite is not a demo of local AI. It brings a frontier experience, AI chat, document summarisation, projects, skills and artefacts, built to run without a server.
A polished, frontier-grade chat experience with fast token-by-token streaming. Stop, regenerate and branch as much as you like, because there is no meter running and no server to wait on.
Import PDFs, Office files and text, and Pathway Lite builds a local vector index. Retrieval augmented answers come with citations, and the index never leaves the disk.
Group chats, documents and instructions into projects. Each project scopes its own retrieval, so your legal folder never bleeds into your recipe experiments.
Build skills in the open Agent Skills format inside Pathway Lite, then deploy the very same skill into frontier systems such as Claude when you need more inference and more power. Your process is portable, not locked in.
Keep your best prompts on hand with fill-in placeholders. Reuse them across chats and projects in two clicks.
Generate documents, code and structured outputs into a slide-in artefact tray, block by block, exactly as Pathway does. Export what you make, keep it on your machine.
A curated catalogue of open source models, downloaded straight from Hugging Face and verified on arrival. Pick the model that fits your Mac's memory, or let the flagship do the work.
The flagship. A reasoning model sized for laptops, quick enough to feel instant on Apple Silicon.
Every download is resumable and integrity-checked with per-file SHA-256 hashes before it can be loaded. Models arrive exactly as their creators published them on Hugging Face, with no modification.
Rewrite the system prompt to shape tone, format and behaviour. The model answers to you, not to a vendor's defaults.
Turn it down for precise, repeatable answers, or up for looser, more creative output. Adjust per chat, see the difference immediately.
Pathway Lite is free, our contribution to community AI. We believe private, on-device intelligence should be available to anyone with the hardware to run it, not gated behind a subscription.
Because inference runs on your own silicon, there is no meter, no token limit and no bill hiding behind the chat box. Updates and new models in the catalogue are included.
No. Conversations, documents, embeddings and settings are stored in a local database on your machine. The only network activity the app performs is downloading models you explicitly choose, and even the app's interface layer is blocked from the network by policy.
Only for the first model download. After that, everything works completely offline, including chat, document search and artefact generation.
Pathway Lite is built for Apple Silicon and runs models through MLX, Apple's machine learning framework. More memory lets you run larger models, and the catalogue tells you what fits your machine before you download.
Yes. You can edit the system prompt, set the temperature, and switch between any downloaded model per chat. Skills in the open Agent Skills format let you teach the model repeatable processes.
Nothing. Pathway Lite is free, our contribution to community AI. Because the models run on your own hardware, there is no usage to meter, no subscription and nothing to bill you for later.
Cloud chatbots run on someone else's computer, which means your prompts, files and conversations travel to their servers and live under their policies. Pathway Lite keeps the entire loop, model, memory and output, on hardware you own.
Download Pathway Lite, pick a model, switch off the Wi-Fi, and see for yourself.
Pathway Lite