AI Model Router
A curated list of open-source and frontier models behind one API. Automatic fallbacks when a provider fails, budgets that actually stop spend, and a log of every request and what it cost.
Free plan: 10,000 requests a month. No card needed to start.
from openai import OpenAI
client = OpenAI(
base_url="https://api.routing.nyuro.ai/v1",
api_key="nyu_live_...",
)
client.chat.completions.create(
model="claude-sonnet-4-6", # or "auto"
messages=[{"role": "user", "content": "Hi"}],
)You choose how each request is routed
Routing follows rules you can read. It picks by price, speed or where the model runs, not by guessing which answer is best.
claude-sonnet-4-6Pin a modelName the model and get exactly that model, never a silent substitute.autoLet the router pickSends each request to the model our published rules match to the task.strategy:costCheapest that fitsAlso strategy:latency for speed and strategy:quality for our top tier.strategy:localKeep it on open modelsRoutes to Luna models on GPUs we run, instead of a vendor API.
Built for teams that ship to customers
- Automatic fallbacks
Where a model has a fallback, a provider error or timeout moves the request to it before your user notices.
- Budgets that stop spend
Set a limit per key and for your whole organization. When it is reached, requests stop instead of running up a bill.
- Every request logged
Model, tokens, cost, speed and which fallback ran, searchable and exportable.
- Per-key model allow-lists
Decide which models each key may call, so a test key cannot reach your most expensive model.
- Audit log
A record of who changed keys, budgets and policies, and when.
- Embeddings included
The same API and key for embeddings, ready for search and RAG.
A curated model list
Every model we list is one we route and test. The catalog shows each one's price.
- OpenAI
- Anthropic
- Google Gemini
- Luna (open source, hosted by us)
- Embeddings
Simple, published pricing
- Vendor models at their list price. No markup on tokens.
- 5.5% fee when you buy credits.
- Bring your own provider keys: 5% fee, no token charge.
- Luna open-source models at our published rate.