Skip to content

Fastlane - Smart LLM Routing

Route every prompt to the best model

A privacy-first LLM router that learns which provider works best for each type of prompt. Save cost while improving quality.

Get Started | View on GitHub


An ONNX ranking model picks the best provider for each request. Simple prompts go to cheap models, complex ones to strong models.

Everything runs on your machine. The ranking model is local. Your prompts only go to the provider you choose.

Route 80% of requests to cheap providers without sacrificing quality. The Pareto optimizer finds the sweet spot.

Drop-in OpenAI-compatible proxy. Works with Cursor, Continue, LangChain, and any tool that speaks the OpenAI API.


Provider Type Notes
OpenAI BYOK gpt-4o, gpt-4o-mini
Anthropic BYOK claude-sonnet-4, claude-3.5-haiku
Groq BYOK llama-3.3-70b, llama-3.1-8b, mixtral, gemma2
OpenRouter BYOK 100+ models via openrouter.ai
Marketplace Pay-as-you-go 2% markup, top up via Razorpay

objective = 0.15 * quality + 0.85 * cost

Prioritize cost savings. Send most requests to cheap providers.

objective = 0.55 * quality + 0.45 * cost

Balance quality and cost equally. Works well for most use cases.

objective = 0.85 * quality + 0.15 * cost

Prioritize quality. Route complex prompts to the best models.


Install Fastlane in 5 minutes: Installation Guide