Router

Pick the right model for the job — then route to it.

📚 Routing is a research field now. 14 verified 2026 papers on how to route — length budgets, session spend caps, prefill-activation signals — and what each suggests we build next.
Routing research →
Balanced
Best overall value — quality with a real eye on cost.
Best quality
Maximize task fit; price is secondary.
Cheapest that passes
Lowest cost above the quality floor.
Fastest that passes
Lowest latency above the quality floor.
Cost-aware quality
Strong quality, but reward cheaper picks.
Providers:
○ No provider keys — routing works; “run live” needs a free key in .env.
Use it as an API
The same decision, callable from anywhere. This is the product you ship.

Describe a job and hit Route. LLM Switchboard classifies the task, applies your policy & constraints, and ranks every model with a transparent score.

Every available model
All sources — Groq · NVIDIA · Cloudflare (cloud), Ollama (local sandbox), and your Railway services. Pick one and Test, or check several and Run on selected to compare them on the same prompt.
ModelSourceContext$/MtokStatus
Loading the directory…
Sign in to continue

LLM Switchboard is private — sign in with Authlee to access the control room.

Sign in with Authlee
← Back to home