Features

• :books: 300+ model catalog — Unified catalog across 20+ providers, sourced from models.dev.
• :dart: Type-safe model IDs — Autocomplete for all cataloged models while accepting arbitrary strings.
• :electric_plug: Provider resolution — Map "provider/model" strings to AI SDK LanguageModel instances.
• :moneybag: Cost calculation — Calculate USD cost from token usage and per-token pricing.
• :package: Subpath imports — Per-provider imports for filtered model lists and typed IDs with zero-bundle overhead.

Install

npm install @funkai/models

Usage

Look up a model

import { model } from "@funkai/models";

const gpt = model("gpt-4.1");

if (gpt) {
  console.log(gpt.name); // "GPT-4.1"
  console.log(gpt.contextWindow); // 1047576
  console.log(gpt.pricing.input); // cost per input token in USD
  console.log(gpt.capabilities); // { reasoning, toolCall, ... }
}

Filter models

import { models } from "@funkai/models";

const reasoning = models((m) => m.capabilities.reasoning);
const vision = models((m) => m.modalities.input.includes("image"));
const cheap = models((m) => m.capabilities.toolCall).toSorted(
  (a, b) => a.pricing.input - b.pricing.input,
);

Resolve providers

import { createProviderRegistry } from "@funkai/models";
import { createOpenAI } from "@ai-sdk/openai";
import { anthropic } from "@ai-sdk/anthropic";

const registry = createProviderRegistry({
  providers: {
    openai: createOpenAI({ apiKey: process.env.OPENAI_API_KEY }),
    anthropic,
  },
});

// Returns a LanguageModel — pass directly to agent()
const lm = registry("openai/gpt-4.1");

Calculate costs

import { model, calculateCost } from "@funkai/models";

const m = model("gpt-4.1");
if (m) {
  const cost = calculateCost(
    {
      inputTokens: 1000,
      outputTokens: 500,
      totalTokens: 1500,
      cacheReadTokens: 0,
      cacheWriteTokens: 0,
      reasoningTokens: 0,
    },
    m.pricing,
  );
  console.log(`Total: $${cost.total.toFixed(6)}`);
}

API

| Export | Description | | ------------------------ | ------------------------------------------------------ | | model(id) | Look up a single model definition by ID | | models(filter?) | Return all models, optionally filtered by a predicate | | MODELS | Complete readonly array of all model definitions | | createProviderRegistry | Create a registry that resolves model IDs to providers | | calculateCost | Calculate USD cost from token usage and pricing |

Subpath Exports

Per-provider subpath imports give access to filtered model lists and typed IDs:

import { openAIModels, openAIModel } from "@funkai/models/openai";
import { anthropicModels } from "@funkai/models/anthropic";

Available for: openai, anthropic, google, google-vertex, mistral, amazon-bedrock, groq, deepseek, xai, cohere, fireworks-ai, togetherai, deepinfra, cerebras, perplexity, openrouter, llama, alibaba, nvidia, huggingface, inception.

Documentation

For comprehensive documentation, see the Models concept and model() reference.

License

MIT