# `LlmCall.Usage`
[🔗](https://github.com/gilbertwong96/llm_call/blob/main/lib/llm_call/usage.ex#L1)

Token accounting as reported by the provider.

`total_tokens` is the provider's total. A provider that reports input split across
counters — Anthropic's `input_tokens` excludes the cached part — is summed by the
codec, so the field stays the real total.

Pricing and cost are outside the core: they depend on provider, model, region,
cache, and tier, none of which is a wire field.

# `t`

```elixir
@type t() :: %LlmCall.Usage{
  cache_read_tokens: tokens(),
  cache_write_tokens: tokens(),
  input_tokens: tokens(),
  output_tokens: tokens(),
  reasoning_tokens: tokens(),
  total_tokens: tokens()
}
```

# `tokens`

```elixir
@type tokens() :: non_neg_integer() | nil
```

---

*Consult [api-reference.md](api-reference.md) for complete listing*
