For providers

Turn idle GPU hours into routed revenue.

Register any OpenAI-compatible endpoint and let the pool route paid developer traffic to you. You set the price floor; Policate handles trust, routing, escrow, and payouts.

83.3%

Provider share

50 cases

Benchmark gate

<3s

Activation p95

Health checkedcontinuous latency probes

Capacity routedregion and concurrency aware

Escrow cleareddaily Stripe payouts

The provider loop

From idle to earning in four steps.

No cloud console, no contracts. Bring the capacity you already have.

Step 1Ollama · vLLM · TGI
Connect capacity
Point the pool at any server that speaks /v1/chat/completions. Expose it publicly with a free Cloudflare Tunnel — no ports, no DNS, no static IP required.
Step 2price · region · concurrency
Set your terms
Choose your model, a per-1M-token price floor, a geographic segment (global / US / EU), and max concurrency. You stay in control of your economics.
Step 350-case benchmark
Serve traffic
The pool runs a 50-case benchmark on registration. Pass-rate ≥ 80% and p95 < 3000ms flips you to active. Then routing, health checks, and quality spot-checks run continuously.
Step 4escrow · Stripe payouts
Get paid
Earnings accrue per token. Credits clear escrow after the dispute window, then pay out to your Stripe account — no invoices, no minimums.

Economics

You keep 83.3% of every token.

A flat 16.7% commission covers routing, trust scoring, escrow, fraud detection, and Stripe payout accounting. No listing fees, no minimums, no lock-in.

0.0%

Commission

Daily

Payout cadence

Listing fee

Revenue split per $1.00

Your payout83.3%

Pool commission16.7%

$0/ mo estimated

1× A100 · qwen2.5-coder-32b · ~62% utilization · $0.74/1M floor

Estimate

Model your monthly payout.

Drag the slider to your expected daily traffic. Payout is your 83.3% share after the 16.7% commission — actual earnings depend on your floor, utilization, and trust score.

Estimate your earnings

Monthly payout after 16.7% commission

Requests per day4,000 / day

Gross / mo

$71.04

Commission (16.7%)

-$11.86

Your payout / mo

$59.18

You keep 83.3% of every token served

Estimate only. Assumes ~800 tokens/request and a $0.74/1M price floor over 30 days. Actual earnings depend on your floor, utilization, model demand, and trust score. Earnings clear escrow after the dispute window, then pay out via Stripe.

Compatibility

If it speaks OpenAI, it can earn.

Any server exposing an OpenAI-compatible chat completions endpoint qualifies.

Ollama

Local model runner. One command to serve open weights.

localhost:11434

vLLM

High-throughput serving with continuous batching.

OpenAI server mode

Any /v1 endpoint

TGI, LiteLLM, llama.cpp server, or your own gateway.

/v1/chat/completions

bash

# 1. Expose your local model server (free Cloudflare Tunnel)
cloudflared tunnel --url http://localhost:11434
# → https://my-gpu-xxxxx.cfargotunnel.com

# 2. Register the provider
policate-pool register \
  --endpoint https://my-gpu-xxxxx.cfargotunnel.com \
  --model qwen2.5-coder-32b \
  --price-floor 0.74 \
  --region eu \
  --max-concurrent 4

Any OpenAI-compatible engine can join the pool

Ollamalocalhost:11434

vLLMOpenAI server mode

LM Studiolocal server

llama.cppserver /v1

TGItext-generation-inference

LocalAIdrop-in /v1

LiteLLMproxy /v1

Any /v1 endpointOpenAI-compatible

Trust & quality

Higher trust means more traffic.

Trust scores (0–100) drive routing priority. They rise with clean spot-checks, strong uptime, and low latency — and protect developers from bad actors.

Quality spot-checks

The pool periodically replays known-answer prompts and grades responses. Consistent quality raises your trust score and routing priority.

Health & latency

Continuous health probes track availability and p95 latency. Degraded providers are pulled from routing until they recover — protecting your score.

Escrow & disputes

Earnings are held in escrow through a dispute window. Fraud detection flags anomalies before funds release, keeping the marketplace honest.

FAQ

Provider questions, answered.

What do I actually earn?

You keep 83.3% of every token served — a flat 16.7% commission covers routing, trust scoring, escrow, fraud detection, and Stripe payout accounting. There are no listing fees, no minimums, and no lock-in. Use the earnings estimator above to model a monthly payout for your traffic.

How do I connect capacity?

Point the pool at any server that speaks /v1/chat/completions — Ollama, vLLM, TGI, llama.cpp, LM Studio, LocalAI, or your own gateway. Expose it publicly with a free Cloudflare Tunnel: no open ports, no DNS, no static IP required.

How does the pool decide I’m ready?

On registration the pool runs a 50-case benchmark. A pass rate ≥ 80% and p95 < 3000ms flips you to active. After that, continuous health probes and quality spot-checks keep your trust score current and drive how much traffic you receive.

When and how do I get paid?

Earnings accrue per token into escrow and clear after a dispute window, during which fraud detection can flag anomalies before funds release. Cleared balances pay out to your Stripe account on a daily cadence — no invoices, no minimums.

Do I stay in control of my pricing?

Yes. You set the per-1M-token price floor, choose your model, pick a geographic segment (global / US / EU), and cap max concurrency. The pool never routes below your floor, and you can update your terms at any time.