pydantic/genai-prices

📊 GenAI Prices – What It Is

GenAI Prices is a small but practical open‑source library that lets developers look up how much it should cost to call a given large‑language‑model (LLM) inference endpoint. It ships:

• Data – a curated, versioned JSON registry of pricing rules for hundreds of models across dozens of providers (OpenAI, Anthropic, Azure, Bedrock, Gemini, etc.). The data includes historic rates, tiered pricing, off‑peak discounts and even per‑token‑type costs.
• Logic – code that matches a user‑supplied provider‑ID and model‑ID to the correct entry, handling quirks like renamed models or provider‑specific pricing schemes.
• Clients – ready‑to‑use packages for Python, JavaScript/TypeScript, and Go, each exposing a tiny API (get_price(...)) and a command‑line tool (genai-prices) for quick look‑ups.
• Extensibility – the repository encourages contributions of new providers or price updates, and it already pulls data from sources such as Helicone, LiteLLM and OpenRouter.

The project is part of the Pydantic Stack, which aims to give developers a type‑safe, production‑ready foundation for AI agents.

---

🔧 Key Features (as listed in the README)

Feature	What It Means
Advanced matching logic	Tries hard to map the exact provider/model you specify to the correct price entry, even when IDs differ slightly.
Historic & changing prices	Stores old rates (e.g., for o1) so you can see how costs evolved over time.
Variable daily pricing	Supports providers that change rates by time of day (e.g., Deepseek off‑peak pricing).
Tiered pricing for Gemini	Handles separate charges for very large context windows.
Price‑discrepancy detection	Tools to compare the registry against other public sources.
Multi‑language packages	Python, JS/TS, and Go libraries + CLIs.
Future API & web UI	Planned, but not yet released.

---

📦 How to Use It

Python

pip install genai-prices

Then in code:

from genai_prices import get_price
price = get_price(provider="openai", model="gpt-4o", tokens=1000)
print(price)  # → {'input': 0.005, 'output': 0.015, 'currency': 'USD'}

The accompanying CLI (genai-prices) mirrors this functionality from the terminal.

JavaScript/TypeScript

npm i @pydantic/genai-prices

import { getPrice } from '@pydantic/genai-prices'
const price = getPrice({ provider: 'anthropic', model: 'claude-3-opus', tokens: 500 })
console.log(price)

A genai-prices command is also provided.

Go

go get github.com/pydantic/genai-prices/go

price, _ := genaiprices.GetPrice("openrouter", "meta-llama/Meta-Llama-3.1-70B-Instruct", 2000)
fmt.Println(price)

---

📂 Data Files

• v2 (current) – prices/new_data/v2/data.json (full) and data_slim.json (free‑model‑stripped, smaller). Schemas are provided.
• v1 – frozen legacy files kept for backward compatibility.

You can download these JSON files directly and embed them in your own tooling if you prefer not to depend on the library.

---

🤝 Contributing & Community

• Submit price updates via pull requests (see prices/README.md).
• Benchmark the price‑lookup performance (benchmarks/README.md).
• If you run a model‑hosting service, you’re encouraged to expose a JSON or API endpoint with your pricing so the project can ingest it.
• Join the Pydantic Slack (badge link in the README) for discussion.

---

⚠️ Accuracy Disclaimer

The authors stress that the numbers are indicative only. Providers rarely publish machine‑readable price tables, so the data is assembled from public docs, community contributions, and other open‑source price lists. Always verify against your actual provider bill.

---

📚 Where It Fits

GenAI Prices fills a niche gap for developers building LLM‑driven applications, cost‑monitoring dashboards, or budgeting tools. By providing a single, type‑safe source of pricing information across the fragmented LLM ecosystem, it helps avoid surprise bills and makes cost‑aware prompting easier.