Pricing and usage

Billing is usage-based and per model. You pay for the input and output tokens of each request at the served model’s published per-million-token rate. There are no seats, no tiers, and no minimums. You are billed only for what you call.

Prices are set per model and published on the pricing page. Rates differ by model and by provider, so always check the current rate for the model id you plan to call.

How a request is priced

Every chat completion returns a usage object. Your cost for that request is computed from those token counts and the model’s rates:

cost_usd = (prompt_tokens / 1,000,000) * input_rate_per_million
         + (completion_tokens / 1,000,000) * output_rate_per_million

The result is rounded to six decimal places (micro-dollar precision), then deducted from your organization’s USD balance.

prompt_tokens are your input tokens (the messages you send), billed at the model’s input rate.
completion_tokens are the model’s output tokens, billed at the model’s output rate.

Input and output are priced separately, so a model can charge a different rate for each. Refer to the pricing page for both rates per model.

The usage object

Every non-streaming completion response includes a usage object. All three fields are always present and are integers.

object

Token accounting for the request.

Show properties

integer

Number of input tokens in the request.

integer

Number of output tokens generated in the response.

integer

Sum of prompt_tokens and completion_tokens.

A typical response body looks like this:

{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "created": 1730000000,
  "model": "claude-sonnet-5",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "Hello!" },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 12,
    "completion_tokens": 8,
    "total_tokens": 20
  }
}

You can read usage directly from the SDK response to track spend in your own app:

from openai import OpenAI

client = OpenAI(
    base_url="https://api.infyrence.com/v1",
    api_key="sk-...",
)

resp = client.chat.completions.create(
    model="claude-sonnet-5",
    messages=[{"role": "user", "content": "Hello!"}],
)

print(resp.usage.prompt_tokens)      # input tokens
print(resp.usage.completion_tokens)  # output tokens
print(resp.usage.total_tokens)       # sum of both

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.infyrence.com/v1",
  apiKey: "sk-...",
});

const resp = await client.chat.completions.create({
  model: "claude-sonnet-5",
  messages: [{ role: "user", content: "Hello!" }],
});

console.log(resp.usage.prompt_tokens);     // input tokens
console.log(resp.usage.completion_tokens); // output tokens
console.log(resp.usage.total_tokens);      // sum of both

curl https://api.infyrence.com/v1/chat/completions \
  -H "Authorization: Bearer sk-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-5",
    "messages": [{ "role": "user", "content": "Hello!" }]
  }'

Streaming responses also carry a usage object on the final chunk, so you get the same accounting whether or not you stream.

Balance and deduction

Your balance is denominated in USD. When a request completes, its cost is deducted from your organization’s balance in a single atomic operation. Concurrent requests cannot race each other, and your balance is clamped at zero, so it can never go negative. When your balance is exhausted, calls to platform-billed models fail with 402 and an error of type insufficient_credits:

{
  "error": {
    "message": "Insufficient balance. Add funds, or use your own provider key (BYOK) to continue.",
    "type": "insufficient_credits",
    "code": null,
    "request_id": "4f0c2e1a-9b3d-4a2e-8c1f-77a0b2d5e6f1"
  }
}

To recover, add funds in the dashboard or switch the request to your own provider key (see BYOK below).

BYOK versus platform balance

You can run requests two ways.

Platform balance

Infyrence calls the provider for you and deducts the request cost from your USD balance at the model’s published rate. This is the default.

Bring your own key (BYOK)

The request runs on your own provider key. Infyrence routes it but does not charge your balance. BYOK requests cost $0 in platform credits.

BYOK is useful when you already hold provider credits or committed-use discounts. You keep those rates and still get one unified endpoint, one Infyrence key, and automatic failover.

Models with no published rate

Billing fails safe. If a requested model is not in the pricing catalog, the request is charged $0 rather than an unknown or default amount. A pricing gap can never overcharge you. Prices are cached briefly on the edge and refresh automatically, so newly listed rates take effect within minutes.

Where to see your spend

Per request, in code

Read the usage object from each response and total it in your application.

In the request logs

Each request is logged with its model, token counts, status, latency, and the USD charged. Use logs to attribute spend to specific calls and API keys.

In the dashboard

View your balance, aggregate usage, and per-model spend in the dashboard.

Next steps

Browse models and rates

List the live catalog with GET /v1/models and look up per-model rates.

Pricing page

See the current per-million-token rate for every model.

Get started

Guides

API reference

How a request is priced

The usage object

Balance and deduction

BYOK versus platform balance

Platform balance

Bring your own key (BYOK)

Models with no published rate

Where to see your spend

Next steps

Browse models and rates

Pricing page

​How a request is priced

​The usage object

​Balance and deduction

​BYOK versus platform balance

Platform balance

Bring your own key (BYOK)

​Models with no published rate

​Where to see your spend

​Next steps

Browse models and rates

Pricing page

How a request is priced

The usage object

Balance and deduction

BYOK versus platform balance

Models with no published rate

Where to see your spend

Next steps