AI model router

The AI model router that chooses by cost, quality, latency and your rules.

An AI model router that decides per request which model - or agent, tool or person - should answer: scored on quality, cost and latency, constrained by your data and region policy, explained in every response. Open source; hosted or self-hosted.

Get a free API key Route this request in the playground

Free plan, no card. Open source under Apache-2.0; self-hosting is free forever.

From the benchmark

9.53 vs 9.15

judged quality, routed vs always-frontier

Blind judges scored the routed answers higher than the frontier model's on the same 150 prompts, while the router spent 89% less. Two use cases went the other way; the page says which.

The whole benchmark, with the cases the router lost

The problem

The right model depends on the request. The request changes every time.

A support reply, a SQL fix, a contract summary and a market analysis want different models - and sometimes not a model at all but a tool, an agent or a person. Choosing once, for the whole application, means being wrong most of the time.

Quality that matters for one request is wasted on the next
Cheap models fail silently on the hard cases
Models, agents and tools are chosen in different places with different rules

How the router answers it

Three things it does for every request.

One pool, every kind of target
Models, agents, skills, tools, prompts and human queues are declared the same way and ranked together. A request that needs doing rather than answering goes to an agent.
Your objective, not ours
Weights for quality, cost, latency and risk per request or per workspace; four presets - Balanced, Cheapest, Best, Fastest - for people who do not want to tune.
Policy before the score
Data boundary, region, tenant, cost ceilings and allowed targets are applied first. A target that fails a rule is not a candidate, and the trace says so.

A real request

Our churn rose from 3.1% to 5.4% while NPS stayed flat. Reason step by step about the causes.

The decision comes back with the chosen target, its confidence, the ranked alternatives and what each would have cost - before you execute anything.

Route it in the playground and read the trace

The callbash

curl -s https://opensmartroute.ai/api/v1/route \
  -H "Authorization: Bearer $OSR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "<prompt>"}'

Where this stands next to the alternatives, and how to go deeper.

Install

npm install @opensmartroute/sdk

Typed SDKs for Python and TypeScript

Or no install at all: the hosted API answers a plain curl with the decision, and the browser extension shows the price and the router's pick under the composer of the AI chat sites you already use.

Route your first request in under a minute.

OpenSmartRoute is open source and the free plan keeps the full trace of every decision. Create a workspace, mint a key and send the request above.

Create a free account Read the benchmark

Free plan, no card. Fifteen thousand decisions a month with the full trace.

The AI model router that chooses by cost, quality, latency and your rules.

The right model depends on the request. The request changes every time.

Three things it does for every request.

One pool, every kind of target

Your objective, not ours

Policy before the score

Our churn rose from 3.1% to 5.4% while NPS stayed flat. Reason step by step about the causes.

Where this stands next to the alternatives, and how to go deeper.

Route your first request in under a minute.

The AI model router that chooses by cost, quality, latency and your rules.

The right model depends on the request. The request changes every time.

Three things it does for every request.

One pool, every kind of target

Your objective, not ours

Policy before the score

Our churn rose from 3.1% to 5.4% while NPS stayed flat. Reason step by step about the causes.

Where this stands next to the alternatives, and how to go deeper.

Route your first request in under a minute.