This page covers the lineup and pricing of Naia's own models (naia-0.9-*). On top of a proven
open base we add a Korean voice · memory · privacy stack — AI you own and run on your own hardware.
Model lineup
| Model | VRAM | Use | Status |
|---|---|---|---|
| naia-0.9-omni-24g | 24GB | Realtime voice (Korean-first, multilingual; multimodal) — real-time voice cascade (STT, LLM, TTS) | Available |
| naia-0.9-coding-24g | 24GB | Coding / agent work (single GPU) | Planned |
| naia-0.9-omni-48g | 48GB | Larger omni — long context, higher-quality voice/multimodal | Planned |
Available right now: naia-0.9-omni-24g. Module details: 4.2 naia-0.9-omni-24g module; free 1-minute trial: 4.3 Live Demo.
How to use
- Offline — run on your own GPU (4.4 Offline version): a $10 / month subscription gives you all Naia models. Pull the container and run it on your own GPU — no cloud, data never leaves your machine, no usage billing. Individuals only (B2B by separate arrangement).
- Online — rented cloud GPU (4.5 Online version, planned): in preparation.
All subscriptions and licenses are personal (individual) licenses — single-user, non-transferable, no redistribution. (B2B by separate arrangement.)
External models such as Gemini are charged at 110% of API cost (details on the Billing & credits page).