Struct ModelPricing 

pub struct ModelPricing {
    pub input_per_million: f64,
    pub output_per_million: f64,
    pub cached_input_per_million: Option<f64>,
    pub cache_write_per_million: Option<f64>,
    pub batch_input_per_million: Option<f64>,
    pub batch_output_per_million: Option<f64>,
    pub flex_input_per_million: Option<f64>,
    pub flex_output_per_million: Option<f64>,
    pub prompt_cache_min_tokens: Option<u32>,
    pub effective_at: DateTime<Utc>,
}

Fields

input_per_million: f64

USD per 1M input tokens.

output_per_million: f64

USD per 1M output tokens.

cached_input_per_million: Option<f64>

USD per 1M cached input tokens (Anthropic 10%, OpenAI 10%, Gemini 10%).

cache_write_per_million: Option<f64>

USD per 1M cache-creation (cache-write) input tokens. Anthropic charges ~1.25× the base input rate for tokens written to the prompt cache. None for providers with no documented write premium (cost path unchanged).

batch_input_per_million: Option<f64>

USD per 1M batch (async) input tokens. Providers with a batch tier (OpenAI / Anthropic / Gemini) bill async requests at ~50% of standard input. None for providers with no batch tier.

batch_output_per_million: Option<f64>

USD per 1M batch (async) output tokens (~50% of standard output). None for providers with no batch tier.

flex_input_per_million: Option<f64>

USD per 1M input tokens under OpenAI’s Flex service tier (service_tier: "flex") — a synchronous-but-slower tier billed at Batch API rates (~50% of standard). None for models/providers with no Flex tier; presence is the eligibility gate (only models that carry a Flex rate may be opted into service_tier=flex). See developers.openai.com/api/docs/guides/flex-processing.

flex_output_per_million: Option<f64>

USD per 1M output tokens under the Flex service tier (~50% of standard output). None when the model has no Flex tier.

prompt_cache_min_tokens: Option<u32>

Provider minimum prefix length, in tokens, before a cache_control breakpoint actually caches (shorter prefixes silently don’t cache). Anthropic varies this by model (2048–4096); None when not documented.

effective_at: DateTime<Utc>

When this pricing took effect (for historical replay).

Implementations

impl ModelPricing

pub fn cache_write_rate_per_million(&self, tier: CacheWriteTier) -> Option<f64>

USD per 1M cache-write (creation) tokens for the given TTL tier.

• FiveMin → the catalog’s cache_write_per_million (the 5-minute/1.25× rate Anthropic applies to bare ephemeral writes).
• OneHour → the documented 2× base-input rate, but only when a 5-min write premium is documented (i.e. the provider tiers cache writes at all). Providers with no write premium return None for both tiers so the caller falls back to the plain input rate, unchanged.

Returns None when no write premium applies, so callers price the remaining tokens at input_per_million.

pub fn flex_eligible(&self) -> bool

Whether this model is eligible for OpenAI’s Flex service tier (service_tier: "flex"). Eligibility is catalog-driven: a model is flex-eligible iff it carries a Flex input rate. OpenAI lists Flex prices only for supported models (gpt-5.x family); o3 / o4-mini are batch-only “specialized models” and therefore carry no Flex rate.

pub fn flex_rates_per_million(&self) -> Option<(f64, f64)>

The Flex (input, output) per-million rates when this model is flex-eligible, else None. Both are present together for an eligible row (the catalog carries the pair); a missing output rate falls back to the standard output rate so a partially-populated row stays priceable.

Trait Implementations

impl Clone for ModelPricing

fn clone(&self) -> ModelPricing

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ModelPricing

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for ModelPricing

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for ModelPricing

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

where __S: Serializer,

Serialize this value into the given Serde serializer.