Locai One is an on-prem AI appliance built for an organisation to run AI inside its own environment.
It combines NVIDIA Blackwell GPU hardware, Locai OS, local model serving, the Locai App and developer API access in one system. It is designed to sit in an office or server environment and serve multiple users over the organisation’s own network.
The standard Locai One has one NVIDIA Blackwell GPU with 96 GB of VRAM. Locai One Pro has two NVIDIA Blackwell GPUs with 192 GB of combined VRAM.
Because inference runs locally, the appliance does not need to send prompts to an external AI model provider. It can be connected to a normal network, used with a direct Ethernet connection, or configured to operate fully air-gapped.