Hardware
Personal AI that lives in your home, on hardware you own. Two editions, one assistant. Early access, with pricing announced soon.
For Mac (Apple silicon)
A quiet, efficient Apple silicon Mac mini running Albena. On-device Apple intelligence plus local models, and ideal for homes.
For NVIDIA RTX
An RTX GPU workstation or server-class machine, with compact AI-box options planned. Bigger local models, vision and camera AI, multi-user households, estates and small businesses.
Compare
Qualitative by design. We will publish tested figures at launch, not before.
| Feature | Hub for Mac | Hub for NVIDIA |
|---|---|---|
| Memory | Unified memory, shared by CPU and GPU. Tiers set at order. | Dedicated GPU memory (VRAM) tiers, plus system RAM. |
| Model size class | Small to mid-size local models, comfortable for a home. | Mid to large local models. Higher VRAM tiers run bigger ones. |
| Rooms and voices | A household and its rooms. | Multi-user households, estates and small businesses. |
| Camera and vision AI | Light, on-demand image understanding. | Always-on, multi-camera vision workloads. |
| Apple on-device intelligence | Yes, alongside local models. | Not applicable. Local models only. |
| Power and noise | Low power draw, very quiet. | Higher power draw. Fans can be audible under load. |
| Form factor | Compact desktop. | Workstation or server class. Compact AI-box options planned. |
| Pricing | Early access pricing, announced soon. | Early access pricing, announced soon. |
Already have a machine
Install Albena on your own Mac or NVIDIA Linux box. Same assistant, same approvals, same controls. You supply the hardware, we supply the software.
Coming later
Small room satellites so you can say “Hey Albena” from the kitchen, the garage or the guest room, and be heard by your Hub. No dates yet. We will announce when they are ready.
Setup
Connect power and your network. Albena Hub starts on its own.
From any phone or laptop on your network, follow the guided setup. Pick your connectors and set your approval rules.
Your assistant is live. Ask her something, and approve what she proposes.
Final contents are confirmed at launch.
Nothing reaches your Hub from outside your home unless you turn it on, and then only the paths you choose.
Updates are verified before they install. If one fails a health check, the Hub returns to the last good version by itself.
One documented reset wipes the device back to a clean state, so you can hand it on or start again.
Memory, conversations and settings live on hardware you own. Cloud models are used only if you allow it.
Early access pricing is announced soon. Tell us which edition you want and we will write first.