⚙️ OpenRouter Auto (model ID: openrouter/auto)

• Launch Date: November 8, 2023
• Context Window: Up to 2 million tokens GitHub+13OpenRouter+13OpenRouter+13OpenRouter+2Medium+2OpenRouter+2
• Routing Logic: Powered by Not Diamond, choosing from dozens of top-tier models to meet prompt requirements OpenRouter+1OpenRouter+1
• Routed Models Include:
  • OpenAI GPT-4o variants (e.g. openai/gpt-4o-2024-08-06, gpt-4o-mini)
  • Anthropic Claude (e.g. claude-3.5-opus, Claude-haiku)
  • Google Gemini Pro & Flash
  • Mistral, LLaMA, Cohere Command R‑Plus, DeepSeek, Quasar Alpha, Kraken, and others https://callin.io/+10OpenRouter+10OpenRouter+10

🧠 Key Technical Specs

1. Auto‑Routing & Fallback

• If your requested model fails or hits a rate limit, openrouter/auto automatically falls back to another model in your routing array. The fallback IDs can be defined manually using the models parameter array YouTube+10OpenRouter+10OpenRouter+10.

2. Uniform API Interface

• Uses the same OpenAI-compatible interface for every call:
  POST https://openrouter.ai/api/v1/chat/completions
• Response includes the actual model used in a model field; billing and billing rates correspond to that specific underlying model OpenRouter+1apidog+1OpenRouter+4OpenRouter+4Relevance AI+4.

3. Pricing Strategy

• You are billed at the same rate as the underlying provider’s model — OpenRouter does not markup inference pricing (though a 5–5.5% fee applies when buying credits or using BYOK) OpenRouter+5OpenRouter+5OpenRouter+5.
• Example models and providers are exposed in the /models endpoint for runtime pricing evaluation apidog+9OpenRouter+9OpenRouter+9.

4. Unified Response Schema

• Full JSON metadata returned includes:
  • usage (token usage)
  • model (actual model used)
  • choices array with completion
  • Optional: structured_outputs, tool call fields, etc. OpenRouter+2OpenRouter+2OpenRouter+2OpenRouter+1OpenRouter+1

📊 Spec Summary Table

🛠️ When to Use OpenRouter Auto

• ✅ If you want best-effort quality and latency, without manually switching models.
• ✅ If you need high availability and uptime fallback, with no single point of failure.
• ✅ If you want cost efficiency by letting OpenRouter select cheaper models when appropriate.
• ✅ For building user‑facing AI chat that seamlessly upgrades to higher‑capability models when needed.

🔧 Example API Request (Python / OpenAI SDK Compatible)

pythonCopyEditfrom openai import OpenAI
client = OpenAI(base_url="https://openrouter.ai/api/v1", api_key="YOUR_KEY")
completion = client.chat.completions.create(
  model="openrouter/auto",
  messages=[{"role": "user", "content": "Explain quantum mechanics simply"}],
  temperature=0.7
)
print(completion)

The response will indicate which underlying model served the request and return usage stats. Relevance AI+2OpenRouter+2Medium+2Medium+13OpenRouter+13OpenRouter+13OpenRouterOpenRouterhttps://callin.io/

⚠️ Developer Tips:

• Use usage: { include: true } to report token usage in the API response.
• If you prefer cost-optimized routing, append the :floor variant (e.g., openrouter/auto:floor) to prioritize cheaper providers.
• To prioritize speed and throughput, use :nitro.
• You can also mix specific models into a fallback chain manually with the models array for deterministic routing overrides. OpenRouter+2OpenRouter+2OpenRouter+2OpenRouter