pied piper
Get API key
REC·PIED-PIPER-NET / live feed
03:34:07:00·24 fps
cam-07 / eu-fra-7c3d   3840×2160   ● on-air
iso 1600·f/1.4·1/48
Live · 3,127 computers online in 64 countries

The most powerful decentralized compute network.

Pied Piper turns thousands of everyday computers into one giant network — putting cheap, private, always-on computing power within reach of anyone. No data centers, no middlemen. Just the world's spare machines, working for you.

Buy compute Earn with your computer
Network status · last 60 s
Computers online
3,127
in 64 countries
Words / sec
184,320
generated right now
AI models ready
47
Llama, DeepSeek, Qwen…
Paid to members · 24h
+8.4%
$44,217
real earnings
Live feed
›fra-7c3dservedllama-3.3-70b·412 ms·◊ 0.0024›iad-1a4bservedqwen2.5-72b·198 ms·◊ 0.0011›sin-9d12serveddeepseek-v3·628 ms·◊ 0.0042›lhr-2e8fservedmixtral-8x22b·301 ms·◊ 0.0019›nrt-5c91servedllama-3.3-70b·142 ms·◊ 0.0008›gru-6b73servedphi-4·88 ms·◊ 0.0003›syd-3d5aservedgemma-2-27b·244 ms·◊ 0.0015›dxb-8a02servedllama-3.1-8b·76 ms·◊ 0.0002›fra-7c3dservedllama-3.3-70b·412 ms·◊ 0.0024›iad-1a4bservedqwen2.5-72b·198 ms·◊ 0.0011›sin-9d12serveddeepseek-v3·628 ms·◊ 0.0042›lhr-2e8fservedmixtral-8x22b·301 ms·◊ 0.0019›nrt-5c91servedllama-3.3-70b·142 ms·◊ 0.0008›gru-6b73servedphi-4·88 ms·◊ 0.0003›syd-3d5aservedgemma-2-27b·244 ms·◊ 0.0015›dxb-8a02servedllama-3.1-8b·76 ms·◊ 0.0002
02Live

Right now, a computer somewhere is doing your work.

LIVEUpdated 2s ago · public stats
Words generated · today
1,842,760,013
+18.4% w/w
Computers online
3,127
across 64 countries
Avg response time
142 ms
as fast as the cloud
Paid to members · 24h
$44,217
real cash
Cost vs big cloud
$0.34
up to 95% cheaper
Active hosts · t = now
3,127 online
Americas894 online
EMEA1,247 online
APAC986 online
Why Pied Piper

Good for buyers. Great for members.

If you need compute

Cheaper, cleaner, private.

  • No data centers
    Runs on real people's computers — not warehouses of servers.
  • Greener by default
    Uses power and machines that are already switched on.
  • Private by design
    Work is spread across the network, so no single party sees it all.
  • The cheapest anywhere
    Up to 95% less than the big cloud providers.
  • Always online
    Thousands of machines, around the clock, worldwide.
Buy compute
If you have a computer

Get paid while you sleep.

  • Cash while you sleep
    Your computer earns whether you're using it or not.
  • Real passive income
    Get paid for hardware you already own.
  • Member perks
    Early members get bonuses, priority, and a say in the network.
  • One click to start
    Just open a tab — no install, no setup, no risk.
  • Cash out anytime
    Withdraw your earnings whenever you want.
Start earning
llama-3.3-70b
qwen2.5-72b
deepseek-v3
mistral-large
mixtral-8x22b
gemma-2-27b
phi-4
command-r-plus
yi-34b
falcon-180b
nemotron-70b
olmo-2-13b
llama-3.3-70b
qwen2.5-72b
deepseek-v3
mistral-large
mixtral-8x22b
gemma-2-27b
phi-4
command-r-plus
yi-34b
falcon-180b
nemotron-70b
olmo-2-13b
01Drop-in

One line. The rest of your stack stays.

Swap your base_url. Keep your SDK, your prompts, your tool-calling, your streaming, your JSON schemas. Pied Piper speaks the OpenAI dialect — chat, embeddings, vision, the lot.

chatcompletionsembeddingsvisiontoolsjson_schemastreaming
No SDK swap
No prompt rewrites
No vendor lock
quickstart.py
1from openai import OpenAI
2
3client = OpenAI(
4    base_url="https://piedpiper.run/v1",
5    api_key=PIEDPIPER_KEY,
6)
7
8stream = client.chat.completions.create(
9    model="llama-3.3-70b",
10    messages=[{"role": "user", "content": "Hello, pied piper."}],
11    stream=True,
12)
13
14for chunk in stream:
15    print(chunk.choices[0].delta.content or "", end="")
stream
142 ms TTFT
03Host

Your idle GPU. A browser tab.

Pied Piper runs entirely in WebGPU. No installer, no admin password, no daemon left behind. Close the tab and you're out — open it again and you're back. Earnings sit in your wallet either way.

  • WebGPU detects your card
    Works on Chrome, Edge, and Brave today. Safari Tech Preview is close.
  • Set your own floor price
    Pied Piper routes around the price. You're paid in ETH on settlement.
  • Attested by random sampling
    Pied Piper re-runs ~1% of your outputs to verify integrity.
  • Stay anonymous by default
    A wallet is the only identity you need.
Start hosting
piedpiper.run/host
Hosting
llama-3.3-70b.q4_k_m
on NVIDIA RTX 4090 · 24GB
Floor price
/ 1M output tok
$0.42market $0.48
Tokens
43,821
Requests
127
Earned
◊ 0.0142
GPU utilization
82%
Recent requestsSTREAMING
POST/v1/chat/completionsllama-3.3-70b412 ms847 tok
POST/v1/chat/completionsllama-3.3-70b198 ms210 tok
POST/v1/embeddingsbge-large24 ms512 tok
POST/v1/chat/completionsllama-3.3-70b628 ms1,204 tok
POST/v1/chat/completionsllama-3.3-70b301 ms492 tok

Two sides of the same pied piper.

For developers
  1. 01
    Point your SDK at Pied Piper
    Set base_url to piedpiper.run/v1 and use your existing API key flow. Anything OpenAI-compatible works — Python, TS, Go, curl.
  2. 02
    Pick a model, send a request
    Browse 47 open-weight models. Pied Piper routes to the fastest available host with capacity and a healthy attestation.
  3. 03
    Pay only for what you use
    No subscriptions, no minimums — usually 60–85% cheaper than the big providers. Top up in seconds.
Read the docs
For hosts
  1. 01
    Open piedpiper.run/host in a tab
    WebGPU detects your card. Pick a model, set your floor price, hit start. No install, no daemon.
  2. 02
    Earn while you sleep
    Your computer answers requests in the background. You stay anonymous, and you can stop anytime.
  3. 03
    Cash out anytime
    You get paid as requests come in. Withdraw whenever you like, or spend your earnings on Pied Piper yourself.
Become a host