Struct Usage 

pub struct Usage {
    pub completion_tokens: usize,
    pub prompt_tokens: usize,
    pub total_tokens: usize,
    pub service_tier: Option<String>,
    pub prompt_audio_seconds: Option<u64>,
    pub num_cached_tokens: Option<u64>,
    pub prompt_tokens_details: Option<PromptTokensDetails>,
}

Token usage returned by Mistral’s chat completions and embeddings endpoints.

See https://docs.mistral.ai/api/ (UsageInfo schema). The three counts are always present; the remaining fields are populated by Mistral on a best-effort basis (e.g. cached-token information appears once a prompt is large enough to be cached).

Fields

completion_tokens: usize

prompt_tokens: usize

total_tokens: usize

service_tier: Option<String>

Capacity tier that served the request, when Mistral reports it.

Although the generated UsageInfo reference currently omits this field, the live chat-completions wire includes values such as "standard" in both blocking responses and terminal stream chunks. Keeping it here prevents the provider-native raw_completion and raw_stream surfaces from silently discarding that wire metadata.

prompt_audio_seconds: Option<u64>

Duration in seconds of audio tokens in the prompt (audio-input models only).

num_cached_tokens: Option<u64>

Total cached prompt tokens reported at the top level. Some Mistral responses populate this in addition to (or instead of) prompt_tokens_details.cached_tokens.

prompt_tokens_details: Option<PromptTokensDetails>

In-depth breakdown of prompt token usage (currently only cached tokens).

Implementations

impl Usage

pub fn cached_tokens(&self) -> u64

Returns the number of cached prompt tokens, preferring the structured prompt_tokens_details.cached_tokens field and falling back to the top-level num_cached_tokens. Returns 0 when neither is present.

pub fn audio_tokens(&self) -> u64

Tokens charged for audio in the prompt. 0 for every non-audio turn.

pub fn input_tokens(&self) -> u64

Every token charged against the prompt.

Mistral reports audio outside prompt_tokens: a Voxtral turn answering a 375-audio-token clip reports prompt_tokens: 6, audio_tokens: 375, completion_tokens: 2 and total_tokens: 383. Counting only prompt_tokens as input leaves input + output short of total by the whole audio payload.

Trait Implementations

impl Clone for Usage

fn clone(&self) -> Usage

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Usage

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl Default for Usage

fn default() -> Usage

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for Usage

fn deserialize<__D>( __deserializer: __D, ) -> Result<Usage, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Display for Usage

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl From<&Usage> for Usage

fn from(usage: &Usage) -> Usage

Converts to this type from the input type.

impl From<Usage> for Usage

fn from(usage: Usage) -> Usage

Converts to this type from the input type.

impl Serialize for Usage

fn serialize<__S>( &self, __serializer: __S, ) -> Result<<__S as Serializer>::Ok, <__S as Serializer>::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.