model-metadata-central

Typed registry of LLM model metadata — context windows, pricing, provider routing — for TypeScript/Node.js projects.

Install

npm install model-metadata-central

Usage

Full registry

import { getAllModels, getAllProviders } from "model-metadata-central";

// Populate a model selector dropdown
const models = getAllModels();
const modelOptions = models.map((m) => ({
  value: m.model_id,
  label: m.model_name,
  contextWindow: m.context_window,
  pricePerMillion: m.cost_per_million_tokens,
}));

Look up a single model

import { getModel } from "model-metadata-central";

const model = getModel("gpt-4o");
if (model) {
  console.log(model.context_window); // 128000
  console.log(model.cost_per_million_tokens); // { input: 5, output: 15 }
}

Filter by provider

import { getModelsByProvider } from "model-metadata-central";

const anthropicModels = getModelsByProvider("anthropic");
const openrouterModels = getModelsByProvider("openrouter");

Provider routing

import { getProviderModelId, getProvider } from "model-metadata-central";

// Get the model ID for a specific provider
const openaiId = getProviderModelId("gpt-4o", "openai"); // "gpt-4o"
const openrouterId = getProviderModelId("gpt-4o", "openrouter"); // "openai/gpt-4o"

// Get provider config (base URL, auth type, etc.)
const provider = getProvider("openai");
console.log(provider.base_url); // "https://api.openai.com/v1"
console.log(provider.auth_type); // "api_key"

Tree-shakeable named exports

Import only the models you need — bundlers will exclude the rest.

import { gpt4o, claudeOpus47 } from "model-metadata-central";

// gpt4o and claudeOpus47 are fully typed ModelMetadata objects
// The full registry + helpers are excluded from the bundle

TypeScript types

All types are exported for your own data structures:

import type { ModelMetadata, ProviderMetadata, TokenCosts } from "model-metadata-central";

function logCost(model: ModelMetadata) {
  if (model.cost_per_million_tokens) {
    const cost = model.cost_per_million_tokens as TokenCosts;
    console.log(`$${cost.input} per 1M input tokens`);
  }
}

API

| Function | Returns | Description | | -------------------------------------------------- | ------------------------------- | --------------------------------- | | getModel(id) | ModelMetadata \| undefined | Lookup by model_id | | getAllModels() | readonly ModelMetadata[] | All models | | getModelsByProvider(providerId) | readonly ModelMetadata[] | Filter by provider | | getModelOnProvider(providerId, providerModelId?) | ModelMetadata \| undefined | Find model on a specific provider | | getProvider(id) | ProviderMetadata \| undefined | Provider config | | getAllProviders() | readonly ProviderMetadata[] | All providers | | getProviderModelId(modelId, providerId) | string \| undefined | Provider-specific model ID |

Data

• Registry compiled from /models/*.yaml and /providers/*.yaml at build time
• 73 models across 18 providers
• Prices are in USD per 1,000,000 tokens

Schema

Matches model-metadata.schema.json and provider.schema.json.