Connect safely.
Connect on BNB Smart Chain for identity only. GPit will not request a signature, approval, transaction, seed phrase, or private key.
Connect your wallet to unlock hardware setup.
INDEPENDENT GPU COMPUTE MARKETPLACE
Sell idle compute.
Rent the grid with your balance.
Put idle GPUs to work across an open compute network. Inspect hardware, benchmark capacity, accept workloads, and track verified compute credits.
BSC MAINNET IDENTITY · ESTIMATED NETWORK VIEW · COMPUTE CREDITS ARE NON-TRANSFERABLE
NODE CONTROL
hardware access · workload routing · verificationwalletdisconnectedconnect or start a guest session
hardwarenot inspectedWebGPU capability unknown
nodeidleinspect hardware to begin
compute balance0.00 CRverified compute credits
Set up your device.
One click checks browser GPU support and completes a local CPU benchmark. Then you can start your node.
not run
Connect your wallet first.
Run one verified job.
Routes a compatible compute workload through the verification pipeline. Only verified work earns compute credits.
- assigned
- compute
- upload
- verify
- credited
Complete hardware setup to enable your node.
Job history.
No jobs yet. Complete the setup to request one.
JOB CENTER
compatible workloads · verified results onlyACTIVE ASSIGNMENTNONE
JOB ID—
MODELNo active job
EST. CREDIT0.00 CR
01 assigned02 compute03 upload04 verify05 settle
Start your node, then choose a compatible workload from the queue.
0s waitingWaiting for compute service · no credits issued
Gemma 4 31B · multimodal31B · high-memory worker
- work units
- 144
- duration
- ~20 sec
- base reward
- 2.40 CR
Qwen3.8 · inferenceopen-weight target · MoE routing
- work units
- 112
- duration
- ~15 sec
- base reward
- 2.05 CR
DeepSeek V4-Pro · reasoninghigh-throughput · score 55+ required
- work units
- 196
- duration
- ~26 sec
- base reward
- 3.85 CR
ESTIMATED GRID TELEMETRY
[ run node ]RTX 4090 · eu-west● EST. ONLINE
util62%
- temp
- 63°C
- power
- 307W
- tok/s
- 67.3
- sessions
- 7
- tokens
- 19,661
- uptime
- 1h 37m
Gemma 4 31B · multimodal inference · google
▸ estimated stream · 425 tok · 44.2s
RX 7900 XTX · us-east● EST. ONLINE
util78%
- temp
- 69°C
- power
- 346W
- tok/s
- 81.9
- sessions
- 11
- tokens
- 28,043
- uptime
- 2h 08m
Qwen3.8 · inference target · qwen
▸ estimated stream · 612 tok · 38.9s
TWO WAYS IN
both sides of the marketYour GPU takes jobs.
Connect an identity, inspect WebGPU, benchmark the device, and route compatible workloads through verification.
- runtimebounded compute workload
- creditverified jobs only
- settlecompute credit ledger
Demand meets supply.
Route inference workloads across available GPUs through one OpenAI-compatible interface.
- endpointOpenAI-compatible API
- routingmodel and region aware
- recordsusage-level observability
OPEN-WEIGHT MODEL NETWORK
target profiles for compatible workersQwen3.8open-weight inference target
MOEPROFILEDeepSeek V4-Proreasoning workload target
REASONINGPROFILEGemma 4 31Bmultimodal workload target
31BPROFILE