One API. Every model you need.

Inference Cloud is the production layer for model inference. One balance, one key, and complete visibility into every token you spend.

Start building Read the docs

Pay-as-you-goPer-token transparencyNo long-term lock-in

inference-cloud / first-request

Live example

Request

cURL

curl https://api.inference.cloud/v1/chat/completions \
  -H "Authorization: Bearer sk_live_xxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "model-pro",
    "messages": [{
      "role": "user",
      "content": "Ship it."
    }]
  '

Response

200 OK

{
  "id": "req_8f2c…",
  "usage": {
    "input_tokens": 842,
     "output_tokens": 372
  },
  "cost": "$0.00384",
 "latency_ms": 642
}

Tokens

1,214

Cost

$0.00384

Latency

642ms

The platform

Infrastructure that gets out of your way.

Everything between your product and the model, handled. The console gives you the details most platforms hide.

Drop-in API compatibility

Use OpenAI-compatible chat completions or the Anthropic-compatible messages API. Change the base URL, keep your tooling.

Every model in one catalog

Compare context windows, capabilities, and per-token pricing. Pick a model without stitching together separate APIs.

Usage you can audit

Per-request tokens, latency, cost, and status. Daily breakdowns and a complete request ledger by default.

Scoped API keys

Create as many keys as your workflow needs. Rotate or revoke instantly, and reveal a secret exactly once.

Simple balance model

Fund a single balance and draw it down per token. Transparent deposits and a running transaction ledger.

Security by default

Cookie-authenticated console, server-side API keys, and sensible error contracts for production integrations.

Works withTypeScript SDKPython SDKcURLOpenAI-compatible clientsAnthropic-compatible clients

How it works

From zero to first token in minutes.

No infrastructure to provision, no minimum commitment. Create an account, pick a model, and send a request.

Create your account

Create an account

Open the console, set up your workspace, and generate your first API token.

Fund your balance

Choose a supported network and token, then send the exact quoted amount.

Point your SDK at us

Set the base URL and model ID. Streaming, tools, and structured output just work.

Pricing

Pay for tokens. Nothing else.

Every model has a published input and output rate per million tokens. Add credits to your balance and draw them down as you build. Cached reads and writes are priced separately where a model supports them.

View model pricing

Per 1M tokens

Per-token rates

Input, output, and cache pricing surfaced before you send a request.

Every call

Request-level cost

Each request records its token counts, cost, latency, and final status.

Prepaid

One balance

Top up once. Use it across every model in the catalog.

Ship inference, not infrastructure.

Create an account and make your first request against a production-ready API today.

Get started free Authentication guide

No card required to create an account.