How it works
Download once. Run forever.
Quenderin turns your phone or laptop into a self-contained AI. There's no server doing the thinking for you — the model itself lives on your device. Setup is three steps, and it only happens once.
- 01
Probe your hardware
On first launch, Quenderin reads what your device actually has — total and free RAM, the chip, and whether there's a usable GPU. It uses that to recommend the largest model your device can run well, not just the largest one that fits. If nothing fits comfortably, it tells you plainly instead of stalling.
- 02
Download a model once
Pick the recommendation or choose your own — anywhere from a 0.4 GB ultra-light build to a 9 GB flagship. The download runs over Wi-Fi and resumes in the background if you switch apps or lose signal, so a big model finishing overnight is no problem. Every download is integrity-checked before it's ever loaded.
- 03
Go offline
From then on, everything runs locally. Before you leave the grid, a clear check confirms the model is present and ready — a green "ready to go offline". After that, airplane mode, a tunnel, or three days on a trail make no difference.
On-device, really
What "on-device" actually means.
Most assistants are a thin app in front of someone else's data center. Quenderin is the opposite: the weights, the inference, and your conversation history all stay on the hardware in your hand.
- No prompt or response is ever sent to a server
- No account, no sign-in, no API keys to manage
- No analytics or telemetry — there's nothing phoning home to disable
- Conversations are stored locally and are yours to delete
- The one and only network call is the initial model download
Right-sized for you
It picks a model that actually runs.
Bigger isn't always better on a phone. Quenderin matches the model to your device's memory headroom, leaving room for the operating system so things stay responsive — and it warns you before a model that would swap, stutter, or fail to load.
The result: a capable assistant that feels instant, instead of a flagship model that crawls. You can always override the recommendation and choose your own trade-off.
Bring your AI offline.
It's open source and runs from GitHub today. Star the repo to follow along — and watch releases to hear the moment the mobile apps land.