What is GenKaKu?
GenKaKu is a decentralized AI inference network. Instead of routing your prompts through corporate data centers, GenKaKu connects you directly to a distributed network of GPU workers — regular people sharing their compute power.
AI powered by people, not data centers.
How it works
You send a message. The orchestrator finds an available worker. The worker runs the model on their GPU and streams tokens back to you in real-time. Your prompts aren't stored, and the worker never sees who you are — it gets the text and nothing else. No corporate filter deciding what you're allowed to ask.
One model
Chat runs on Qwen3.8 27B Uncensored (qwen3.8-27b-uncensored) — one model for the whole network, so there's nothing to pick. It's uncensored, and it does web search, vision (send it an image), and an extended thinking mode.
| Model | Cost | Where it runs | Notes |
|---|---|---|---|
| Qwen3.8 27B Uncensored | about 1 credit a message | Workers across the distributed network | uncensored; web search + tools, image input, thinking mode |
Image generation runs on its own pool of image workers.
Credits
Inference is paid for with credits. A credit is $0.001, and a dollar buys 500 of them. You don't need any token to use GenKaKu.
- Sign in and you get free credits every day. A plan gives you more
- Top up with USDC for anything past the day's credits; text is billed per token
- Workers earn 70% of the USD value of the credits spent on jobs they complete, paid in USDC
The stack
- Workers are desktop apps that run on your machine and serve inference via Ollama with hardware acceleration — CUDA (NVIDIA), Metal (Apple Silicon), or Vulkan (AMD/Intel)
- The orchestrator is a Socket.io server that handles job routing, worker matching, and real-time token streaming
Why?
Centralized AI providers censor their models, log your prompts, and can revoke access at any time. GenKaKu is the alternative — private, uncensored, and owned by no one.
Anyone can use GenKaKu or contribute compute and start earning.