pricetoken 0.4.0

Real-time LLM pricing data — typed client, cost calculator, and static pricing

pricetoken

Real-time LLM pricing data — typed client, cost calculator, and static pricing.

Python SDK for PriceToken. Zero runtime dependencies.

Install

pip install pricetoken
# or
uv add pricetoken

Quick Start

Calculate cost offline (no API call)

from pricetoken import calculate_model_cost

cost = calculate_model_cost("claude-sonnet-4-6", 1_000_000, 100_000)
print(cost.total_cost)  # $4.50

Fetch live pricing

from pricetoken import PriceTokenClient

client = PriceTokenClient()
pricing = client.get_pricing()

for model in pricing:
    print(f"{model.display_name}: ${model.input_per_m_tok}/MTok in, ${model.output_per_m_tok}/MTok out")

With API key (higher rate limits)

client = PriceTokenClient(api_key="pt_your_key_here")

Cost Calculation

from pricetoken import calculate_cost, calculate_model_cost

# By model ID (uses bundled static pricing)
cost = calculate_model_cost("gpt-4.1", 500_000, 100_000)
print(cost.input_cost)   # $1.00
print(cost.output_cost)  # $0.80
print(cost.total_cost)   # $1.80

# With explicit rates
cost = calculate_cost("custom-model", 5.0, 15.0, 1_000_000, 500_000)
print(cost.total_cost)   # $12.50

Static Pricing Data

21 models from Anthropic, OpenAI, Google, and DeepSeek are bundled for offline use:

from pricetoken import STATIC_PRICING

for model in STATIC_PRICING:
    print(f"{model.model_id}: ${model.input_per_m_tok}/MTok")

API Reference

Method	API Path
get_pricing(provider=, currency=)	GET /api/v1/pricing
get_model(model_id, currency=)	GET /api/v1/pricing/{model_id}
get_history(days=, model_id=, provider=)	GET /api/v1/pricing/history
get_providers()	GET /api/v1/pricing/providers
compare(model_ids, currency=)	GET /api/v1/pricing/compare
get_cheapest(provider=, currency=)	GET /api/v1/pricing/cheapest

Types

All types are dataclass(slots=True) for performance and type safety:

• ModelPricing — model pricing data
• CostEstimate — cost calculation result
• PriceHistoryPoint — historical price data point
• ModelHistory — price history for a model
• ProviderSummary — provider overview with stats
• PriceTokenError — API error (extends Exception)

Type aliases: ModelStatus, DataConfidence, Source (all Literal types).

Requirements

• Python >= 3.10
• Zero runtime dependencies

License

MIT