modelpricing-ai 2026.2.3

Python client for ModelPricing.ai cost estimates

modelpricing-ai

Python client for the ModelPricing.ai API.

Installation

pip install modelpricing-ai

Usage (sync, typed)

from modelpricing_ai import ModelPricingClient

client = ModelPricingClient(api_key="YOUR_API_KEY")

estimate = client.estimate(
    model="gpt-4o-mini",
    tokens_in=125000,
    tokens_out=42000,
    trace_id={"requestId": "abc-123"},
)
print(estimate.total)  # total USD cost

• Base URL: https://api.modelpricing.ai
• Auth: Authorization: Bearer <API_KEY>
• Endpoint: POST /v1/estimate
• Request body:

{
    "metrics": { "tokensIn": 1000, "tokensOut": 2000 },
    "model": "<model-name>",
    "traceId": { "...": "..." }
}

Error handling

• 401 Unauthorized: invalid or missing API key
• 422 Unprocessable Entity: validation error (e.g., invalid model or metrics)
• Other 4xx/5xx: raised as a generic server error

Retry behavior

The client automatically retries on transient errors with exponential backoff:

• Retries: 5xx server errors and network/connection errors
• No retry: 4xx client errors (401, 404, 422)
• Default: 3 attempts with exponential backoff (1s, 2s, 4s, up to 10s max)

Configure retry behavior:

# Customize max retries (default is 3)
client = ModelPricingClient(api_key="YOUR_API_KEY", max_retries=5)

# Disable retries
client = ModelPricingClient(api_key="YOUR_API_KEY", max_retries=1)

This handles transient infrastructure issues like Cloudflare errors, DNS failures, and temporary service disruptions.

License

MIT

Async usage (typed)

import asyncio
from modelpricing_ai import AsyncModelPricingClient

async def main():
    client = AsyncModelPricingClient(api_key="YOUR_API_KEY")
    estimate = await client.estimate(
        model="gpt-4o-mini",
        tokens_in=125000,
        tokens_out=42000,
        trace_id={"requestId": "abc-123"},
    )
    print(estimate.total)  # total USD cost

asyncio.run(main())