Model Routing

AssistantRouter lets you choose from multiple AI models across providers and configure fallback behavior.

Available Models

AssistantRouter supports models from multiple providers:

Explicit Model Selection

Specify a model directly in your request:

const response = await client.chat.completions.create({
  model: 'claude-3-haiku-20240307',
  messages: [{ role: 'user', content: 'Hello!' }],
});

Model Fallbacks

Configure a fallback model on your assistant so that if the primary model is unavailable, requests automatically retry with the fallback:

const assistant = await fetch('https://api.assistantrouter.com/v1/assistants', {
  method: 'POST',
  headers: {
    'Authorization': `Bearer ${apiKey}`,
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    name: 'Reliable Assistant',
    model: 'claude-3-5-sonnet-20241022',
    model_config: {
      fallback_models: ['gpt-4o', 'gemini-1.5-pro'],
      max_retries: 3,
    },
  }),
}).then(r => r.json());

Model Pricing

Costs vary by model and are deducted from your wallet. Prices are sourced dynamically from OpenRouter at the time of each request.

Best Practices

1. Start cheap - Use faster/cheaper models (Gemini 1.5 Flash, Claude 3 Haiku) for simple tasks
2. Use fallbacks - Always configure backup models
3. Monitor costs - Track spending in the dashboard
4. Test models - Different models excel at different tasks
5. Consider latency - Smaller models are faster