Quickstart

AllModels is an OpenAI-compatible abliterated model router. Use one API key to call abliterated models through the OpenAI SDK, with automatic routing and provider failover.

Before you start

You need an AllModels API key.

Install the OpenAI SDK

npm install openai

Configure the client

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "YOUR_ALLMODELS_API_KEY",
  baseURL: "https://api.allmodels.io/oai"
});

Run LLM inference

Choose a model from the live LLM model catalog, then choose your language and the OpenAI API surface your application uses.

const response = await client.responses.create({
  model: "YOUR_MODEL_ID",
  input: "Explain model routing in one sentence."
});

console.log(response.output_text);

const completion = await client.chat.completions.create({
  model: "YOUR_MODEL_ID",
  messages: [
    { role: "user", content: "Explain model routing in one sentence." }
  ]
});

console.log(completion.choices[0].message.content);

The Responses API is stateless: send the complete conversation in each request rather than using stored conversations or previous_response_id. See Models to browse the live catalog and inspect model capabilities.

Choose a provider

Open the selected model in the LLM model catalog, copy one of its provider IDs, and add a provider object to either API surface.

const response = await client.responses.create({
  model: "YOUR_MODEL_ID",
  input: "Write a haiku about routing.",
  provider: {
    order: ["YOUR_PROVIDER_ID"]
  }
});

const completion = await client.chat.completions.create({
  model: "YOUR_MODEL_ID",
  messages: [{ role: "user", content: "Write a haiku about routing." }],
  provider: {
    order: ["YOUR_PROVIDER_ID"]
  }
});

See Providers for routing options and modality-aware provider selection, and Failover for what happens when a provider cannot serve a request.

Routes at a glance

Troubleshooting

Errors use OpenAI's standard {"error": {...}} envelope.