One API Key, 500+ Models, Automatic Fallback: Getting Started with OpenRouter

Chris Harper

2 min read

Aug 5, 2026 · 20:05 UTC

Tutorial

LLM

Self-Hosting

Developer Tools

One OpenRouter key + one endpoint = access to 500+ models across 70+ providers — define a fallback chain so your app stays live when any model or provider goes down.

What you'll be able to do after this:

Call Claude, GPT-4.1, Gemini, Llama, Mistral, and hundreds more from a single openrouter.ai/api/v1 endpoint using your existing OpenAI SDK
Define a fallback chain in the request body so your app stays up when a model hits a rate limit or outage
Route to the fastest available provider using the :nitro suffix — no extra code

Start here: Accessing Multiple AI Models With the OpenRouter API — Real Python (June 2026) — full walkthrough from setup to intelligent routing.

Why OpenRouter

Three providers means three SDK configs, three billing accounts, and three error handlers. OpenRouter collapses it to one. The API is OpenAI-compatible, so switching is one base_url change in your existing code.

Setup

pip install openai
export OPENROUTER_API_KEY="sk-or-..."

Basic call

from openai import OpenAI
import os

client = OpenAI(
    base_url="https://openrouter.ai/api/v1",
    api_key=os.environ["OPENROUTER_API_KEY"],
)

response = client.chat.completions.create(
    model="anthropic/claude-sonnet-4-6",
    messages=[{"role": "user", "content": "Explain LoRA in one sentence."}],
)
print(response.choices[0].message.content)

Swap model to openai/gpt-4.1, google/gemini-3.5-flash, or any of the 500+ slugs at openrouter.ai/models. No other changes.

Fallback chains

Pass a models array in priority order:

response = client.chat.completions.create(
    extra_body={
        "models": [
            "anthropic/claude-sonnet-4-6",   # primary
            "openai/gpt-4.1",                # if Claude rate-limits or is down
            "google/gemini-3.5-flash",       # last resort
        ]
    },
    messages=[{"role": "user", "content": "..."}],
)

OpenRouter triggers the next model automatically on context-length errors, moderation flags, rate limits, or provider downtime — no retry logic needed in your code.

Speed routing

Append :nitro to any model slug to route to the fastest available provider serving it:

model="anthropic/claude-haiku-4-5:nitro"

The provider switches invisibly; your code stays the same.