Struct ModelPrice 

pub struct ModelPrice {
    pub input_per_mtok: f64,
    pub output_per_mtok: f64,
    pub cache_read_per_mtok: f64,
    pub cache_write_5m_per_mtok: f64,
    pub cache_write_1h_per_mtok: f64,
    pub fast_multiplier: f64,
    pub us_only_multiplier: f64,
}

Per-million-token prices for one model, in US dollars.

Fields

input_per_mtok: f64

Input (prompt) tokens, dollars per million.

output_per_mtok: f64

Output (completion) tokens, dollars per million.

cache_read_per_mtok: f64

Prompt tokens read from the cache (a cache hit or refresh), dollars per million.

cache_write_5m_per_mtok: f64

Prompt tokens written to the 5-minute cache, dollars per million.

cache_write_1h_per_mtok: f64

Prompt tokens written to the 1-hour cache, dollars per million.

fast_multiplier: f64

What fast mode (usage.speed: "fast") multiplies every rate by, cache rates included; 1 where the model has no fast mode (a request asking for it runs, and is billed, at standard speed).

us_only_multiplier: f64

What US-only inference (inference_geo: "us") multiplies every rate by: 1.1 on Claude 4.6 and later models, 1 where the model has no such premium.

Implementations

impl ModelPrice

pub const fn standard(input_per_mtok: f64, output_per_mtok: f64) -> Self

A model priced with the provider’s standard cache multipliers: a cache read at 0.1× input, a 5-minute write at 1.25×, a 1-hour write at 2×, and no fast mode.

pub fn cost_usd(&self, prompt_tokens: u64, completion_tokens: u64) -> f64

Dollar cost of one round-trip’s token counts.

Cached prompt tokens are billed at the full input rate here: the provider-reported discount varies per provider and per cache tier, and over-reporting cost is the safe direction for a cap (a budget that stops slightly early never overspends). Named rather than silently assumed — see the per-turn-cost-usage-accounting ledger row’s note. A recorded session’s spend is Self::recorded_cost_usd.

pub fn recorded_cost_usd( &self, tokens: &RecordedTokens, fast: bool, us_only: bool, ) -> f64

What a harness’s recorded use cost at these list rates: uncached input, output, cache reads and each cache write at its own rate (the counts are disjoint, as Claude’s usage reports them), all multiplied in fast mode and for US-only inference (the multipliers stack).

Trait Implementations

impl Clone for ModelPrice

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for ModelPrice

impl Debug for ModelPrice

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

Formats the value using the given formatter.

impl PartialEq for ModelPrice

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for ModelPrice