● TAIWAN-BUILT · GLOBAL COMPUTE NETWORK

One API.
Every available GPU.

Route AI inference across sovereign Taiwan infrastructure, official model APIs, and a worldwide community of idle GPUs.

Start building free →Explore live models ↗

✓ No credit card required✓ 500 Green per verified node✓ OpenAI compatible

api.tokens2tokens.com/v1LIVE

REQUESTqwen/qwen3-8b

SMART ROUTERTW · 22ms · verified

COMPUTECommunity GPU

from openai import OpenAI

client = OpenAI({
  base_url="https://api.tokens2tokens.com/v1",
  api_key="t2t_••••••••"
)

response = client.chat.completions.create({
  model="qwen/qwen3-8b",
  messages=[{"role": "user", "content": "Hello Taiwan!"}]
)

GoldOfficial compute

IndigoEarned compute

GreenNetwork credit

How credits work →

✦ EVERYTHING YOU NEED

Compute that moves where demand is.

Open-source nodes, real-time model discovery, and an accountable ledger make idle hardware useful without changing your AI SDK.

⌘

One compatible API

Use OpenAI clients and switch models with one parameter. Streaming, usage, and errors follow familiar shapes.

⑂

Smart marketplace routing

Filter by price, latency, country, verified entity, privacy tier, and quantization.

◫

Live model discovery

Ollama, LM Studio, llama.cpp, and vLLM catalogs appear automatically when nodes connect.

◇

Sovereign when it matters

Gold routes sensitive workloads to contracted official infrastructure in Taiwan.

↗

Earn where you spend

Providers retain 95% in Indigo and can immediately use it for community inference.

⎋

Open provider client

Transparent Rust and Tauri software for Windows, Ubuntu, and macOS.

⌁ OPEN MARKETPLACE

Compute has a market now.

Every live endpoint competes on price, latency, uptime, location, and trust. Token2Token finds the right route for each request.

Browse the model marketplace →

LIVE ROUTE CANDIDATES3 ONLINE

Taiwan OfficialTaiwan · verified

officialGold only18 ms

Formosa LabTaiwan · community

communityAll credits31 ms

Indigo Node 7Japan · community

communityAll credits47 ms

Balanced routingprice × latency × uptime

GPU

HAVE AN IDLE GPU?

Your next inference can pay for itself.

Install the open-source client, connect your existing engine, choose prices and an earnings cap, then join the network.

Become a provider →View source ↗

TOKEN2TOKEN BETA

Put every GPU within reach.

Share GPU. Earn Indigo. Run any model.

Create account →Read the docs

One API.Every available GPU.