Nitro Variants: Prioritize Speed ⚡

The :nitro suffix now works universally across all models! Instead of appearing as separate endpoints, :nitro is a shortcut that sorts providers by throughput (tokens per second). This allows you to always get the fastest available provider for your chosen model.

Example using Nitro shortcut:

{
  "model": "meta-llama/llama-3.1-70b-instruct:nitro",
  "messages": [{ "role": "user", "content": "Hello" }]
}

This is equivalent to setting provider.sort to "throughput" in the API.

Floor Price: Prioritize Cost 💰

Want to minimize costs? The :floor shortcut ensures OpenRouter always selects the lowest-priced provider for your model.

Example using Floor shortcut:

{
  "model": "meta-llama/llama-3.1-70b-instruct:floor",
  "messages": [{ "role": "user", "content": "Hello" }]
}

This is equivalent to setting provider.sort to "price" in the API.

Advanced Provider Sorting

If you want more explicit control over provider selection, you can set the sort field in provider preferences. This disables OpenRouter’s default load balancing and orders providers based on your chosen attribute:

• "price": Prioritize the lowest-cost provider
• "throughput": Prioritize the fastest provider (highest tokens/sec)
• "latency": Prioritize the lowest-latency provider

Example with explicit sorting:

{
  "model": "meta-llama/llama-3.1-70b-instruct",
  "messages": [{ "role": "user", "content": "Hello" }],
  "provider": {
    "sort": "throughput"
  }
}

What’s Next?

• Improved Throughput Metrics: We’re shifting from averages to median throughput, which means speed ratings will better reflect real-world performance.

Get ready for faster and more cost-effective AI access with these new shortcuts! 🎉