Join the TibiOS Network and Community

Build a TibiBox. Join the network. TibiOS runs across a distributed network of affordable computers — you don't need a datacenter to contribute compute. Choose a compatible machine, install TibiOS, and become part of the network.

USE FRONTIER AI MODELS FOR FREE, WITH NO TOKEN LIMITS!!

Build the TibiOS Fleet — NVIDIA Jetson Orin Nano Super 8GB and Orange Pi 6 Plus, two TibiBox options for one distributed AI network.

Illustration for reference only — see exact specifications below.

Where can I buy the hardware?

We don't sell the hardware — yet. An official TibiOS store is planned for late October. Until then, these are recommended configurations and links to retailers so you can build your own TibiBox. Store links are search links and may vary by availability and market — they are not a guarantee that they point to one specific listing. Prices and availability are set by the retailer and may change.

No payment now — we'll email you when it's ready to buy.

The more reservations, the lower the price per device.

🇪🇸 Spain

🌎 Latin America

More countries — Panama, Guatemala — coming next.

🇺🇸 United States

🌍 Other regions

Choose your TibiBox

🟢 TibiBox GPU — NVIDIA Jetson Orin Nano Super, 8 GB

Recommended for: GPU compute, CUDA, AI inference, model execution.

  • 8 GB LPDDR5
  • 67 INT8 TOPS
  • 1024 CUDA cores + 32 Tensor cores
  • 102 GB/s memory bandwidth
  • 7–25 W power envelope

~$249 USD (NVIDIA reference price, Developer Kit)

See Where to buy for links by country.

🔵 TibiBox Network — Orange Pi 6 Plus, 16 GB

Recommended for: CPU, memory, network and distributed workloads.

  • 16/32/64 GB LPDDR5
  • 12-core ARM CPU
  • NPU (NeuralONE), ~30 INT8 TOPS
  • 2 × 5GbE + 2 × M.2 NVMe PCIe 4.0 ×4
  • Ubuntu / Debian

Price varies by memory configuration and retailer.

See Where to buy for links by country.

Don't know which one to choose?

Orin Nano Super Orange Pi 6 Plus
GPU / CUDA⭐⭐⭐⭐⭐Arm Immortalis-G720 MC10
CPU⭐⭐⭐⭐⭐⭐⭐⭐
RAM8 GB16/32/64 GB
Network1GbE2 × 5GbE
AI acceleratorCUDA / Tensor, 67 INT8 TOPSNPU, ~30 INT8 TOPS
Distributed TibiOS⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐
Best forAI computeMemory + networking

Both are useful. TibiOS is heterogeneous by design.

Why different machines?

The TibiOS network doesn't require every node to be identical. Different machines contribute different resources — TibiOS schedules workloads according to GPU capacity, memory, performance and network locality. A heterogeneous fleet can become a distributed AI system, and it's exactly what lets us experiment with heterogeneous scheduling, capacity, locality and, later, Expert placement.

Orin (GPU) + Orange Pi (CPU/Network) TibiOS Distributed AI

Build your node

1

Choose your hardware

Select an Orin, Orange Pi, or another supported TibiOS node.

2

Install TibiOS

Follow the installation guide and configure your machine.

3

Connect your node

Connect it to your network and let TibiOS discover the node.

4

Contribute compute

Your machine becomes part of the distributed TibiOS infrastructure.

Recommended starter fleet: 1 × Orin + 1 × Orange Pi — the smallest heterogeneous setup to start experimenting with distributed AI.

1 2 4 8 16 32+ nodes

Build the network with us.

One machine is an experiment. Thousands of machines become infrastructure.

Build a TibiBox Run TibiOS locally now