Implementation:Helicone Helicone CostOf

Knowledge Sources	Helicone
Domains	Cost Calculation, LLM Observability, Pricing Registry
Last Updated	2026-02-14 00:00 GMT

Overview

Concrete function for looking up per-token cost rates for a model/provider pair, provided by the @helicone/cost package.

Description

The costOf function accepts a model name and provider identifier and returns the cost rate object (ModelRow["cost"]) for that combination, or null if no matching rate is found. The function performs a case-insensitive lookup through two stages:

1. Provider resolution: Finds the provider entry by comparing the provider argument against both the provider name string (exact match) and the pattern regex of each registered provider.
2. Model resolution: Within the matched provider's costs array, finds the cost entry whose model identifier matches using the entry's operator: equals (case-insensitive exact match), startsWith (lowercase prefix match), or includes (lowercase substring match).

If the model string is falsy (null, undefined, or empty), the function returns null immediately.

Usage

Use this function when you need raw per-token cost rates without computing a total cost. It is the building block used by costOfPrompt and the ClickHouse SQL generator.

Code Reference

Source Location

• Repository: Helicone
• File: packages/cost/index.ts (lines 23-59)

Signature

export function costOf({
  model,
  provider,
}: {
  model: string;
  provider: string;
}): ModelRow["cost"] | null

Import

import { costOf } from "@helicone/cost";

I/O Contract

Inputs

Name	Type	Required	Description
model	string	Yes	The model identifier (e.g., "gpt-4o", "claude-sonnet-4-20250514"). Matched case-insensitively.
provider	string	Yes	The provider name (e.g., "OPENAI", "ANTHROPIC") or a provider URL that will be tested against registered patterns.

Outputs

Name	Type	Description
return	ModelRow["cost"] ¦ null	The cost rate object containing per-token prices, or null if no match is found.

Cost Rate Object Fields

Field	Type	Description
prompt_token	number	Cost per prompt/input token in USD
completion_token	number	Cost per completion/output token in USD
prompt_cache_write_token	number ¦ undefined	Cost per cache write token (Anthropic: sum of 5m + 1h)
prompt_cache_read_token	number ¦ undefined	Cost per cache read token
prompt_audio_token	number ¦ undefined	Cost per input audio token
completion_audio_token	number ¦ undefined	Cost per output audio token
per_image	number ¦ undefined	Cost per image
per_call	number ¦ undefined	Fixed cost per API call
prompt_cache_creation_5m	number ¦ undefined	Anthropic 5-minute cache creation cost per token
prompt_cache_creation_1h	number ¦ undefined	Anthropic 1-hour cache creation cost per token

Usage Examples

Basic Usage

import { costOf } from "@helicone/cost";

const rates = costOf({
  model: "gpt-4o",
  provider: "OPENAI",
});

if (rates) {
  console.log("Prompt token cost:", rates.prompt_token);
  console.log("Completion token cost:", rates.completion_token);
}

Provider URL Matching

// The provider argument can also be a URL
const rates = costOf({
  model: "claude-sonnet-4-20250514",
  provider: "https://api.anthropic.com",
});

if (rates) {
  console.log("Has cache pricing:", !!rates.prompt_cache_write_token);
}

Null Handling

const rates = costOf({
  model: "unknown-model-xyz",
  provider: "OPENAI",
});

// Returns null when model is not found in the provider's cost table
if (!rates) {
  console.log("No pricing data available for this model");
}

Related Pages

Implements Principle

• Principle:Helicone_Helicone_Cost_Rate_Lookup

Uses Heuristic

• Heuristic:Helicone_Helicone_Cost_Precision_Multiplier