Quickstart

tokenpricing answers two questions: what does a model cost per million tokens and what will this workload cost — for 3,000+ AI models, in any currency, including prompt-cache rates.

Install

Python

pip install tokenpricing

TypeScript

npm install @atenareply/tokenpricing

The Python package ships the tokenpricing CLI; the npm package exposes the same APIs for Node.js 18+.

Look up a price

tokenpricing pricing openai/gpt-5.2

openai/gpt-5.2 pricing (USD):
  Input per 1M tokens: 1.250000 USD
  Output per 1M tokens: 10.000000 USD
  Cache read per 1M tokens: 0.125000 USD

Add --currency EUR for converted rates (cached FX), or --json for machine-readable output.

Compute a workload cost

tokenpricing cost openai/gpt-5.2 --in 250000 --out 40000 --cache-read 100000 --currency EUR

Cache flags are optional and default to 0.

From code

Python

from tokenpricing import get_pricing_sync, compute_cost_sync

pricing = get_pricing_sync("openai/gpt-5.2", currency="EUR")
print(f"Input: €{pricing.input_per_million:.2f}/1M tokens")

cost = compute_cost_sync(
    "openai/gpt-5.2",
    input_tokens=1000,
    output_tokens=500,
    cache_read_tokens=250,
)

TypeScript

import { getPricing, computeCost } from "@atenareply/tokenpricing";

const pricing = await getPricing("openai/gpt-5.2", "EUR");

const cost = await computeCost("openai/gpt-5.2", 1000, 500, "USD", {
  cacheReadTokens: 250,
});

Next steps

• CLI reference — every flag, JSON shapes, model-id suggestions
• SDKs — async APIs, caching behavior, currency handling
• Webhook notifications — alerts for price moves and model lifecycle events
• Canonical database — consume the raw JSON directly