Benchmark-driven routing
How each call goes to the model that did best on similar past calls.
Production-grade AI
Weflayr’s gateway makes every LLM call reliable, and easy to control & optimize.Prompt caching optimization
The best caching for each model, applied for you, without locking a conversation on a weak model.
Routing strategies
Stay in control: choose the models, prompt caching rules, quality objective and per-customer rules.
Self-heal
Before each call, unsupported parameters, like temperature or a reasoning effort, are removed or mapped to the closest one, so the provider never refuses them.
Fallback and model health
A failed call moves on to the model’s other providers and regions, or to the next model, and unhealthy deployments are tried last.
How it works
- Gateway with Benchmark-Driven Router
- Gateway only
1
Use the gateway, or just log your calls with telemetry, a CSV or via API
Use the Weflayr gateway to call any model of any provider, import past calls from a CSV or via API or use our telemetry instrumentor.
2
Benchmark
Benchmark your calls on dozens of models and define if you want to optimize cost or quality.
3
Route
Create a routing strategy that sends each call to the best model, based on your benchmarks, fallback strategy and/or customer cost rules.
4
Analyze
See your AI cost per feature and your margin per customer once your revenue is in.
5
Repeat
New models come out and your traffic changes: benchmark your recent calls again, and update your routing strategy from the dashboard, with no code to change.
Quickstart
Send your first calls, benchmark them and route them, in a few minutes.
Go further
Platform
Benchmark, route and analyze your calls from the dashboard.
SDKs
Python and Node.js libraries.
API reference
The gateway and the management API.
How it works
The technical details behind routing, evaluation, caching and self-heal.