TibiOS
Collective power, global intelligence.
TibiOS is an open ecosystem where every device contributes compute โ and everyone gains access to more powerful AI. No data centers, no subscriptions: your laptop, your PC and your phone become the network.
TibiOS Core
Routes every request to the right device, by capability
Our reference fleet runs on machines anyone can buy โ Macs with Apple Silicon, an NVIDIA Jetson Orin Nano for GPU compute and an Orange Pi 6 Plus as the network’s memory-rich node.
The first video was generated by AI running across our distributed TibiOS fleet โ a Mac Air, an Orange Pi and an NVIDIA Orin Nano working as one.
Our first model trained by its own fleet: five home machines generated their own training data, trained a LoRA adapter on Qwen3-4B and deployed it back as a new inference lane โ zero cloud, zero token spend.
Base model. MLX on Apple Silicon, Unsloth + TRL on CUDA.
LoRA adapter, rank 8 โ just 0.74% of the model.
Adapter size, served by the head as the @tuned lane.
Talk to the fleet from any browser at tibios.ai.
Add TibiOSBot to a group and the fleet answers there.
OpenAI-compatible endpoint for your own tools.
Use the fleet as a backend for coding agents.
Our mascot โ a little mogwai with big ears who celebrates every new device that joins the fleet. He is hungry for compute.