01
Your key, your session
Every account gets its own API keys and its own sessions. Prompts travel on your key, to your model — nothing is commingled with the rest of the fleet.
Private AI, priced by the token
The intimacy of an at-home server, powered anywhere in the world.
What Infra is
Infra runs a private, OpenAI-compatible API on dedicated GPU hardware. Create an account, issue your own API key and pick a model from the public catalog. You pay published per-token rates for the tokens you use — no subscription, and no sessions shared with anyone else.
Why Infra
01
Every account gets its own API keys and its own sessions. Prompts travel on your key, to your model — nothing is commingled with the rest of the fleet.
02
Every model has a published price per million tokens — input, output and cached input. You see the rate before you send a request, and the bill matches it.
03
Point a single OpenAI-compatible endpoint at the whole catalog. Switch models by changing one line — no new integrations to set up.
Sign up with an email and a password. The console opens right away.
Create a key in the console, and rotate or revoke it whenever you want.
Paste the endpoint and your key into anything that speaks the OpenAI request format.
Models & pricing
These rates are live from the fleet, in USD per 1M tokens. Full catalog lists every model, latency and the cached-input rate.
| Model | Context | Input | Output |
|---|---|---|---|
| deepseek-v4.1-flash-1017 | 1,048,576 tokens | $0.345 | $1.38 |
| glm-5.3-flash | 1,048,576 tokens | $0.172 | $0.575 |
Coming soon
Above the per-token fleet we are building dedicated GPU-hour tiers — reserve a machine for your session alone, billed by the hour, from interruptible economy cards to premium 96 GB hardware. They will appear on this page the day they go live.
Get started
Create an account, issue a key and send your first request today — no invitation needed.