Open LLM inference · on a marketplace of real GPUs

Intelligence on demand, powered by GPUs everywhere.

AI demand is growing faster than the world can build for it — yet powerful GPUs sit idle everywhere. OpenHosst Grid puts them to work, serving open models through one OpenAI-compatible API. More capacity for AI, from hardware that already exists — and a lighter footprint for the planet.

Get your API key Browse live models

Drop-in OpenAI API • Automatic failover • Pay per token • No subscriptions

1 API

OpenAI
compatible

100%

Open
source

$0

Subscriptions
or minimums

Live on the grid

Real models, real-time pricing.

Llama 3.2, Qwen 2.5, Mistral, Phi and more are being served right now by providers around the world. You see live availability and the cheapest node — platform fee included — before a single token is spent.

llama3.2

Meta Llama 3.2 — fast, compact instruction model.

qwen2.5

Alibaba Qwen 2.5 — strong multilingual reasoning.

mistral

Mistral — efficient open-weight workhorse.

Explore the full marketplace

For builders

From base URL to first token in three steps.

01

Create your key

Sign up, generate a consumer key (ohk_…) and top up your wallet. No GPUs to provision, no infrastructure to run.

02

Point your SDK here

Swap your OpenAI base URL for the grid endpoint. Every OpenAI-compatible client, SDK and framework works unchanged.

03

Pay per token

We route to the best available GPU and fail over automatically if one drops. You're billed only for completed requests — nothing else.

quickstart.py

from openai import OpenAI

client = OpenAI(
    base_url="https://grid.openhosst.com/v1",
    api_key="ohk_your_key_here",
)

resp = client.chat.completions.create(
    model="llama3.2",
    messages=[{"role": "user", "content": "Hello, grid!"}],
    stream=True,
)
for chunk in resp:
    print(chunk.choices[0].delta.content or "", end="")

For providers

Turn idle GPU time into income.

Run the open-source daemon beside your local Ollama and the grid feeds it paid jobs. Set your own prices, stay private behind an outbound-only connection, and cash out with Stripe.

terminal

# 1. Install Ollama + pull a model
ollama pull llama3.2

# 2. Run the OpenHosst provider daemon
export OPENHOSST_PROVIDER_KEY=ohp_your_key
npx @openhosst/provider

Start earning