What is Maestro

There are thousands of AI models. Each one is best at something. One writes the best plan. A different one writes the best code. Another makes the best pictures. They do not cost the same, and the best one today may not be the best one next month, because new models arrive every week.

So every task now needs routing. Which model should do it? Which cloud should run it? Did it come back right? If it did not, who gets it next? A business is thousands of tasks. Routing each one to the right model, on the right cloud, and making sure it came back right, has become a skill of its own. Maestro does it.

#What Maestro does

Give Maestro a task. It routes the task to the model that does it best. It puts that model on the cloud that runs it fastest for the least money. It checks what comes back. If the result is not good enough, it tries again, or gives the task to a stronger model. Then it hands the result to the next model, with everything that model needs to know.

Maestro does not do the work. Maestro knows who should, and knows when the work is done right. A maestro does not play every instrument. A maestro knows which one to call on, when to call on it, and hears when a note is wrong.

#Two decisions, two names

#Routing is a skill. We taught a model to do it.

We watched millions of real tasks go to real models. We saw which model did each kind of task best, where it ran fastest, and what it cost. We saw which plans worked, and how many tries each one took. Every route, every check and every result went into one model: Maestro.

#What you get

• The best of every model, in one place.
• The best cloud for every job, without ever opening a cloud.
• One bill: what the models cost, plus 5%.

A router sends one request to one model. Maestro looks at the whole job: which model for each task, which cloud, whether the result passed, and what to do next when it did not. Then it learns from the answer, so the next job goes better.

The Maestro page on zbstack.xyz shows it at work, one step at a time.