Your GPU sits idle.
Someone's model is waiting.
GpuCluster is decentralized GPU compute on Robinhood Chain. Connect your GPU, serve LLM inference for the network, and earn verified compute credits for every token you help generate.
Three steps between your GPU and getting paid.
No pool operator, no order book, no idle hardware. The node client handles scheduling, proof generation, and settlement.
Connect
Install the node client, point it at your GPU, and register it on Robinhood Chain. Your hardware is benchmarked once and listed by real throughput, not advertised specs.
Serve inference
The scheduler routes LLM inference requests to your node when it's idle. You set the ceiling: which models you'll run, how much VRAM to allocate, and when.
Earn verified credits
Every completed job is checked with a proof-of-inference and settled as a compute credit on-chain — no invoices, no trust required, no waiting on a payout batch.
Bring the GPU you already have.
Consumer and datacenter cards both earn. Throughput is measured, not assumed — your credit rate reflects what your card actually delivers on the reference model set.
Every job settles with a proof, not a promise.
When your node finishes a job, GpuCluster generates a proof-of-inference: a signed record of the model, the token count, and the wall-clock time. Robinhood Chain verifies the proof and mints the credit directly to your node's wallet — visible on-chain, disputable by no one.
The chain exists to keep the network honest.
GpuCluster doesn't run the jobs itself — it coordinates a network of GPU owners and verifies their work.
Verified, not assumed
Proof-of-inference checks output against the request before any credit is minted, so payment tracks real work, not uptime.
Open to any node
No allowlist or vendor lock-in. Any GPU that passes the benchmark can join the network and start taking jobs the same day.
Settlement without a middleman
Credits mint straight to your node's wallet on Robinhood Chain. No payout batches, no minimum balance, no platform holding period.
Turn idle silicon into income.
Install the node client, connect your GPU, and start serving inference in under ten minutes.