Community-powered AI computeBeta

Share your GPU. Power the AI network.

Put your GPU to work when you are not using it. Spin-Up AI Worker turns spare hardware capacity into a productive asset, rewarding qualifying completed jobs with AI Credits while helping provide accessible, affordable compute across the distributed AI network.

Join the worker network See how it works Download AI Worker for Windows

Beta notice: Spin-up AI Worker and the distributed network are currently in active testing. Features, compatibility, credits and availability may change, and service interruptions or defects should be expected.

DISTRIBUTED JOB / AI INFERENCE Running
Workflow requestAI work enters the network
Network orchestratorMatches work to available capacity
ACTIVE
Worker A
Worker B
Worker C
Completed AI resultReturned to the workflow
READY
— Active AI worker nodes
— Workloads processed

Loading live network activity…

Your schedule
Live performance
Earn usage credits
Worker identity
How it works

Like a marketplace for GPU capacity

Hardware owners make spare compute available. Spin-Up AI matches that capacity with AI jobs, measures completed work and credits contributors—while customers gain another route to run AI workflows without relying only on high-cost centralized APIs.

1

Install AI Worker

Connect a compatible Windows machine running local AI models through Ollama.

2

Set availability

Choose when your machine may accept work and pause it whenever you need your hardware.

3

Receive a job

The network matches supported model work with an available worker and leases the assignment.

4

Process locally

Your local model performs inference and reports the result and performance measurements.

5

Earn credits

Qualifying completed jobs can earn credits for your own future distributed AI calls.

For contributors

Put idle hardware to productive use

AI Worker gives you local control over availability, models and participation while it connects your machine to the wider Spin-Up AI network.

Use the GPU you already own

Advertise available GPU memory, supported models and processing performance to the network.

Choose your schedule

Define weekly availability and exceptions so contribution happens on your terms.

Contribute and earn

Earn usage credits from qualifying jobs and spend them on future distributed AI calls.

Stay in control

Pause, resume or drain the worker and see its current jobs and network status from the local dashboard.

Manage local models

Discover installed Ollama models, pull supported models and test local inference before accepting jobs.

Measure every job

Track token counts, response time and tokens per second for completed inference work.

For AI users

A broader, lower-cost compute option

A public network of contributed machines can expand available capacity and reduce dependence on premium centralized API calls for suitable AI workflow tasks.

Lower-cost inference

Use community-provided capacity for supported work where a premium API is unnecessary.

Public capacity

Draw from a growing network rather than depending on compute from a single provider.

Elastic workflows

Route compatible workflow steps to available workers as network capacity changes.

Distributed execution

Combine multiple AI jobs into broader workflows coordinated through Spin-Up AI.

For developers

Call it live or work with the queue

Both options use the same inference endpoint. Live mode holds the connection briefly for a completed response; queued mode returns immediately so your application can check it later.

Live

Wait for the result

Set waitSeconds to keep the request open. If work finishes in time, the response includes the generated content.

cURL
curl -X POST \
  "{API_BASE_URL}/api/v1/inference?waitSeconds=30" \
  -H "Authorization: Bearer sk-aiw_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemma4:12b",
    "input": "What is the capital of Australia?"
  }'

# Completed within 30 seconds:
# { "requestId": 10, "status": "Completed",
#   "response": { "content": "Canberra" } }
Queued

Submit now, collect later

Omit waitSeconds for an immediate queue response, then retrieve the request by its returned ID.

cURL
# 1. Submit the request
curl -X POST \
  "{API_BASE_URL}/api/v1/inference" \
  -H "Authorization: Bearer sk-aiw_..." \
  -H "Content-Type: application/json" \
  -d '{"model":"gemma4:12b","input":"Explain async/await"}'

# Response: { "requestId": 11, "status": "Queued" }

# 2. Retrieve its current status or result
curl "{API_BASE_URL}/api/v1/inference/11" \
  -H "Authorization: Bearer sk-aiw_..."
Choose your API hostReplace {API_BASE_URL} with your Spin-Up API address, such as your local or published API host.
Safe fallbackIf live mode reaches its wait limit first, use the returned requestId to retrieve the result later.
Token usage includedCompleted responses report prompt, output and total token counts for billing visibility.

Turn spare compute into AI capacity.

Join the community-powered network, earn credits from qualifying work and help make distributed AI more affordable.

Usage credits are not money or investments. Monetary contributor rewards are planned for a future programme and are not currently guaranteed.

Join the network