Free AI, powered by the crowd
Chat with open models for free — served by GPUs worldwide. Contribute compute, earn credits. No signup.
Prompts are processed by distributed GPU nodes. Don't share passwords or API keys.
Why distributed inference?
Models run across multiple peers. If one goes down, requests automatically route to another — no downtime.
Contribute idle compute and earn credits. Spend them to run larger models across the network. A fair exchange.
Drop-in OpenAI-compatible API. Switch from any provider with one env var. Your tools, your choice.
Run your own swarm or federate with trusted peers. Control membership, models, and data with cross-network routing.
Apache 2.0 licensed. Audit the code, fork it, extend it. Every layer is transparent — no black boxes, no hidden costs.
Every new seeder makes the network faster and more capable. More contributors means more models and lower latency.
Drop-in OpenAI replacement
Change one env var. Everything else stays the same.
Sensitive Data Guard
Outgoing prompts are scanned on-device for API keys, passwords, and PII. High-severity matches automatically route to your local model — sensitive data never leaves your device.
Turn your iPhone or iPad into a node on the network. An M-series iPad Pro is a desktop-lite powerhouse — it runs 3B+ models at 30+ tokens/sec on Metal, serving the swarm, not just using it.
Or route from your phone to a trusted device on your own network — your Mac mini, MacBook Air, or Studio — and run far bigger models without the cloud.
