A sovereign AI handheld. Large language models run entirely on your device — zero cloud, nothing to phone home, no subscription. Your data never leaves your hand.
One principle — your AI, on your silicon, with open firmware. The Core is fast and quiet. The Pro adds a discrete AI accelerator for bigger, sharper models.
Built to be owned, not leased. Open-sovereign silicon, verifiable privacy, hardware you can trust by inspection.
Rockchip RK3588S — no CUDA, no proprietary stack, full Linux mainline. You inspect what runs on your device.
LUKS2 full-disk encryption with TPM-bound keys + a hardened secure element. Even we can't read your data.
Every core feature works without internet. No cloud fallback, ever.
Swap Qwen, Llama, Mistral, Gemma, Phi anytime. Your device, your pick.
Peer-to-peer between your NGARi devices — decentralized AI, no server needed.
Long-lifecycle silicon, replaceable battery, years of software updates.
We're building NGARi ONE on open, sovereign silicon — this is a real product in development, not vapor. Pre-orders reserve your place at production pricing and fund the build. Delivery happens in batches through 2027.
Pre-order ◆ refundable in full anytime before shipment ◆ ships 2027 in batches ◆ +$25 US shipping ◆ questions: support@ngari.ai
NGARi is raising to bring the first sovereign AI handheld to life. If you believe intelligence belongs on hardware you own — this is your moment to be early. Early supporters and investors are the ones who get there first.
Investment opportunity now open. Early backers help fund the first production batch — and get in before the rest of the world hears about it.
Invest in NGARiThis is an investment. All terms, disclosures, and risks are on the campaign page. This page is not an offer to sell securities.
It is a real product we are actively building on open silicon (RK3588S + Hailo-10H). This page is an honest pre-order: we fund the build from your reservations and ship in batches through 2027. Cancel anytime before shipment for a full refund.
Sovereign AI means you own it. We price the hardware honestly and give you open, on-device AI that works forever. There is no mandatory subscription for core AI — the device is complete without it.
Yes. Every core AI feature runs entirely on the device. You control when and whether it connects at all.
Yes — the device supports open models: Qwen, Llama, Mistral, Gemma, Phi and more. It's your silicon; choose the weights you trust.
Pre-orders ship in batches through 2027. Delivery estimate is shown at checkout (US: +$25). International availability and fees are confirmed at checkout.