Permute beta is live on Robinhood testnet:
https://t.co/jLYpfdCQ37
Featuring chat (free + paid tiers), a GPU marketplace (with one GPU deployed for testnet), staking, and USDG‑backed escrows.
All smart contracts are published and verifiable on Robinhood testnet 👇
To solve the proof-of-compute problem (when a provider claims a high-end GPU but runs a cheap laptop CPU) the orchestrator runs SLA profiling and tracks each worker’s tokens per second (TPS). If throughput falls below our benchmark threshold, we derank that worker and stop routing jobs to it.
Soon: ZK proofs or hardware trusted execution environments (TEEs) to verify the exact chip that ran the code.
Permute beta is live on Robinhood testnet:
https://t.co/jLYpfdCQ37
Featuring chat (free + paid tiers), a GPU marketplace (with one GPU deployed for testnet), staking, and USDG‑backed escrows.
All smart contracts are published and verifiable on Robinhood testnet 👇
To become a provider, install the script: npm install -g permute-worker
and ensure your inference engine is running (we suppport ollama and vLLM). Our orchestrator handles the network and data routing using our websocket
Read more: https://t.co/F2mynjH1h9
Permute gives every developers API to discover compute, submit workloads, track jobs, and receive on-chain receipts.
Connect your wallet.
Get your API key.
Start building.
Read more in in our github docs [https://t.co/F2mynjH1h9]
At the core, Permute routes to provider GPUs first for inference (our first workload feature) and jobs. When the capacity is maxed out, we fallback to Groq @GroqLLC for speed and low latency execution
What happens when compute becomes a shared resource?
A machine can become a provider. Agents and developers can bring workloads. Decentralized participants collectively coordinate the system.
That's Permute.
Compute shouldn't belong to a handful of cloud providers. What’s missing is an open network that lets people supply compute and earn, use compute, and participate in the economy built around it.