> ## Documentation Index
> Fetch the complete documentation index at: https://docs.profy.cn/llms.txt
> Use this file to discover all available pages before exploring further.

# Billing Units

> Six billing methods: Token, per-call, per-second, per-10K-character, and more

Profy offers six billing methods for different types of AI capabilities. Each model is bound to a specific billing unit in the platform's pricing catalog, and you can see the applicable billing method when selecting a model.

## Six Billing Methods

<Tabs>
  <Tab title="Token Billing">
    **Billing unit:** `token_split`

    **Use case:** LLM conversational AI

    **How it works:** Input tokens and output tokens are calculated separately, each multiplied by their respective rate, then summed. Different models have different input/output rates.

    A token is the basic unit that language models use to process text. One Chinese character typically corresponds to 1–2 tokens, while one English word typically corresponds to 1–4 tokens. The longer the conversation and the more context involved, the more tokens are consumed.

    <Tip>You can view the number of tokens and credits consumed after each conversation turn. Choosing a lighter model can significantly reduce consumption.</Tip>
  </Tab>

  <Tab title="Per-Call Billing">
    **Billing unit:** `per_call`

    **Use case:** Image generation, music generation

    **How it works:** Each generation request is charged a fixed amount of credits, regardless of the complexity of the generated content.

    For example, generating one image deducts a fixed number of credits, regardless of image size or detail level.
  </Tab>

  <Tab title="Per-Character Billing">
    **Billing unit:** `per_character`

    **Use case:** Text processing tasks

    **How it works:** Charged based on the number of characters processed. Each character is converted to credits at the corresponding rate.
  </Tab>

  <Tab title="Per-10K-Character Billing">
    **Billing unit:** `per_ten_thousand_characters`

    **Use case:** TTS (Text-to-Speech)

    **How it works:** Charged per 10,000 characters. Chinese characters count as 2 characters, while English/numeric characters count as 1.

    <Note>This billing method aligns with upstream providers' "price per 10K characters" pricing model, avoiding rounding errors that would occur with extremely small per-character rates.</Note>
  </Tab>

  <Tab title="Duration Billing">
    **Billing unit:** `duration_second`

    **Use case:** Video generation, speech recognition (ASR)

    **How it works:** Charged based on the duration (in seconds) of the generated or processed audio/video. The longer the duration, the higher the consumption.
  </Tab>
</Tabs>

## Relationship Between Billing Units and Models

Each model in the platform's pricing catalog is assigned a specific billing unit and corresponding rate. This means:

* **Rates vary among similar models**: Even among LLM conversation models, premium and lightweight models have significantly different token rates.
* **Similar capabilities may use different billing methods**: For example, text-to-speech is billed per 10K characters, while speech recognition is billed by duration.

You can view the billing method and estimated consumption for any model when making your selection.

## Rate Reference

| Billing Method            | Formula                              | Influencing Factors               |
| ------------------------- | ------------------------------------ | --------------------------------- |
| Token Billing             | Input credits + Output credits       | Conversation length, context size |
| Per-Call Billing          | Fixed credits × number of calls      | Number of generation requests     |
| Per-Character Billing     | Character count × rate per character | Amount of text processed          |
| Per-10K-Character Billing | (Character count ÷ 10,000) × rate    | Amount of text synthesized        |
| Duration Billing          | Seconds × rate per second            | Audio/video duration              |

<Warning>
  Specific rates are dynamically configured by the platform based on upstream model costs and vary by model. Please refer to the real-time display in the model selector within the product — exact values are not listed here.
</Warning>

## How to Save Credits

<CardGroup cols={2}>
  <Card title="Choose the Right Model" icon="sliders">
    Not every task requires the most powerful model. Use a lightweight model for simple Q\&A and switch to a premium model for complex reasoning.
  </Card>

  <Card title="Keep Context Concise" icon="compress">
    With token billing, longer conversations cost more. Starting a new conversation when appropriate can reduce unnecessary context overhead.
  </Card>

  <Card title="Monitor Consumption" icon="chart-line">
    Credits consumed are displayed after each conversation turn, helping you understand the cost of different operations.
  </Card>

  <Card title="Upgrade Your Plan" icon="arrow-up">
    Paid plans include generous credit bonuses and unlock additional features. See [Subscription Plans](/en/documentation/billing/subscription-plans).
  </Card>
</CardGroup>