Every Kalea is local, private, embodied, and voice-first. Those things are never a paid upgrade — they're the whole point. The tiers differ in how much capability.
Best for tutors, learners & family companions
8GB · 64GB eMMC · blue
Lite + multi-user, multi-model, agents & swarms
16GB · 256GB eMMC · black
Ships late October to mid-November; $30 shipping & handling per order. Both models share the same current platform — the difference is memory (8GB vs 16GB), storage, and how many models Kalea can keep warm at once. Kalea is model-agnostic and built to trade up: the hardware is the vessel, and the intelligence keeps improving as better small models arrive.
Not warmth of personality — both have that. Warmth of models: how many minds she can hold ready at once. That's what 16GB actually buys.
| Kalea lite | Kalea Pro | |
|---|---|---|
| Best for | Tutors, single learners, family companions | Lite + multi-user households, MMO, dev, agents & swarms |
| Memory | 8GB LPDDR5 | 16GB LPDDR5 |
| Storage | 64GB eMMC | 256GB eMMC |
| Platform | Orange Pi 5 Pro · RK3588S 8 core CPU | Orange Pi 5 Pro · RK3588S 8 core CPU |
| Personality Deck | Custom 12-bit ADC PCBA | Custom 12-bit ADC PCBA |
| Models kept warm | One at a time | Fast + deep brain together · both brains resident |
| Brain switching | A reload — about 45 seconds | Instant |
| Small Council (multi-model) | Available, slower | Prewarmed — fast |
| Household swarm | Will join the graph as one sealed node | Built for multi-node, multi-user households — runs the director / distributed Council |
| Kalea Console | Included — on-device guardian surface | Included — on-device guardian surface |
| Privacy & ownership | Full — never a paid upgrade | Full — never a paid upgrade |
Whichever Kalea you bring home, the core stays constant: she runs on the device, your data stays in the room, there's no account and no ad model, and you own her outright. The difference between lite and Pro is memory, storage, speed, and how many minds she can hold at once.