Model Registry & Pricing

easy5 minmodel-registry.ts

You want to pick a model by capability and price — tool support, vision, context size — and estimate what a run will cost.

What it shows

Look up a model and its pricing with getModel() and getPrice()
Filter with listModels()
Refresh the data from LiteLLM with initializeModels(), falling back to the built-in list offline

The code

model-registry.ts

import { getModel, getModelRegistry, getPrice, initializeModels, listModels, shutdownModels } from '@cogitator-ai/models';

const claude = getModel('claude-sonnet-5-5');
console.log(`${claude?.displayName}: ${claude?.contextWindow} tokens, $${claude?.pricing.input}/M input`);

const price = getPrice('gpt-6-luna');
if (price) {
  const cost = (50_000 / 1_000_000) * price.input + (10_000 / 1_000_000) * price.output;
  console.log(`50k in + 10k out on GPT-6 Luna: $${cost.toFixed(4)}`);
}

const cheapToolModels = listModels({
  supportsTools: true,
  maxPricePerMillion: 2,
  excludeDeprecated: true,
});
for (const model of cheapToolModels.slice(0, 5)) {
  console.log(`  ${model.id} (${model.provider})`);
}

const visionModels = listModels({ supportsVision: true, minContextWindow: 200_000, excludeDeprecated: true });
console.log('Vision models with 200k+ context:', visionModels.map((model) => model.id).join(', '));

try {
  await initializeModels();
  console.log('Models after fetching LiteLLM data:', getModelRegistry().getModelCount());
} catch {
  console.log('Offline: using the built-in model list');
}

shutdownModels();

Run it

install · bash

pnpm add @cogitator-ai/models

run · bash

npx tsx model-registry.ts

Or run the full example from a clone of the repository (keys in .env):

repository · bash

npx tsx examples/core/13-model-registry.ts

Model Registry & Pricing

What it shows

The code

Run it

Learn more