> ## Documentation Index
> Fetch the complete documentation index at: https://docs.021labs.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# TokenUsage

> Token usage information from LLM calls

## Definition

```python theme={null}
class TokenUsage(CamelModel):
    prompt_tokens: int
    completion_tokens: int
    total_tokens: int
```

## Fields

<ParamField path="prompt_tokens" type="integer" required>
  Number of tokens in the input/prompt.
</ParamField>

<ParamField path="completion_tokens" type="integer" required>
  Number of tokens in the output/completion.
</ParamField>

<ParamField path="total_tokens" type="integer" required>
  Total tokens used (`prompt_tokens + completion_tokens`).
</ParamField>

## Example

```python theme={null}
from rdk.models import TokenUsage

usage = TokenUsage(
    prompt_tokens=150,
    completion_tokens=75,
    total_tokens=225
)
```

## Automatic Capture

RDK automatically captures token usage from LLM providers:

```python theme={null}
from anthropic import Anthropic
from rdk import observe

@observe(name="chat")
def chat(message: str):
    client = Anthropic()
    response = client.messages.create(
        model="claude-sonnet-4-6",
        max_tokens=1024,
        messages=[{"role": "user", "content": message}]
    )
    # Token usage is automatically captured in the span
    return response.content[0].text
```

## Provider Support

| Provider      | Token Usage                    |
| ------------- | ------------------------------ |
| Anthropic     | Yes                            |
| OpenAI        | Yes                            |
| Google Gemini | Partial                        |
| LangChain     | Depends on underlying provider |

## Cost Estimation

Use token counts to estimate costs:

```python theme={null}
# Example cost calculation (rates vary by model)
COST_PER_1K_INPUT = 0.003   # $3 per 1M input tokens
COST_PER_1K_OUTPUT = 0.015  # $15 per 1M output tokens

def estimate_cost(usage: TokenUsage) -> float:
    input_cost = (usage.prompt_tokens / 1000) * COST_PER_1K_INPUT
    output_cost = (usage.completion_tokens / 1000) * COST_PER_1K_OUTPUT
    return input_cost + output_cost
```

## JSON Serialization

```json theme={null}
{
  "promptTokens": 150,
  "completionTokens": 75,
  "totalTokens": 225
}
```

## See Also

* [Span](/api-reference/models/span) - Contains token usage