Meet the NeoTeam Model Router
Every task is different. Your intelligence should be, too.
One model shouldn't have to do every job. Your team draws on different models for code, research, images and video, without you managing the switches.
Iris · ResearchTask matched40 competitor sites. One clear picture.
Reading at scale needs efficient intelligence. Iris uses a low-cost model to turn forty competitor sites into a market brief.
You can see every choice and change any of them.
How it chooses
The model fits the work. The work gets checked.
There are millions of models, and each is best at something different. The leader for code is not the leader for images, and a low-cost model does most routine work as well as an expensive one. The router keeps capability, cost and quality in view throughout a job.
Understand the task. Bulk reading, complex code, an image, or a customer reply? The task determines the intelligence it needs.
Choose the fit. An efficient model for routine work, a frontier model for hard reasoning, and specialists for images and video. Chosen at the moment the task starts, from any lab, open or closed.
Check the result. Evaluators test the work. A task that fails its check moves to a stronger model for another pass, in the middle of the job.
Keep you in control. The choices and costs stay visible. Inspect what happened, or pick a different model yourself. When a better model appears, the router can use it without you changing tools.
How it produces better work and lower cost
Better work and lower cost come from the same decisions.
| Part of the job | What the Model Router does | Effect on the work | Effect on the cost |
|---|---|---|---|
| Bulk reading, summaries, templated pages | Uses the cheapest model that can do it, often an open model on NeoTeam's own compute. | The same. | Far lower. |
| Hard reasoning, code across many files, final judgements | Uses a frontier model. | Better than a mid-priced model would manage. | The premium is paid only here. |
| Images, video, voice | Uses the model that leads that category. | Better than the tool built into a general assistant. | Paid per use, with no extra subscription. |
| A result that fails its check | An evaluator moves the task to a stronger model. | Weak work is caught before you see it. | A small extra cost, only when it is needed. |
| A request from a user of your live product | Picks the model for that request. | Each user gets the model that suits the request. | The cost of serving each customer falls. |
| A new model is released | It joins the pool and the choices update. | Your team and your product improve without you doing anything. | Your prices fall as the field's prices fall. |
The same idea, for the people who don't write code
OpenRouter is for developers. NeoTeam's Model Router is for everyone else.
Developers switch to the right model for every task with OpenRouter. The NeoTeam Model Router does the same for founders who don't write code, inside a team that finishes the work.
| Dimension | OpenRouter | NeoTeam Model Router |
|---|---|---|
| Who it is for | Developers. | Everyone else: founders and small teams who don't write code. |
| How you use it | In code, through an interface for programmers. | In plain language, by telling Neo what you want. |
| Who picks the model | You do, from hundreds of models, or you turn on its automatic mode. | The router does, task by task, in real time. You can switch any choice. |
| What comes back | A model's answer to one request. | Finished work: a product, a campaign, a set of books, a running company. |
| What you must know | Which model to ask for. | Nothing about models. |
| How you pay | Credits, with model prices passed through plus a fee. | Credits. Model usage at cost plus a 5% fee, shown on every quote. |
The router doesn't stop at launch
Built into the team. Built into what you build.
When your product goes live, the same routing serves your users' requests. Everyday requests get efficient intelligence. Harder ones get a more capable model. The cost of serving each customer falls the way your build cost did.
- Your customers
A question. A request. A new possibility.
- Model Router
The intelligence for this request.
- Your product on NeoCompute
Working for your users. Growing with your company.
Where the router sits
Three connected layers.
NeoStack
brings your people, agents, and tools together.
What it holdsThe founder, her human colleagues, Neo and the AI team, and the tools they work in. Every task a colleague does passes through the router.
Model Router
puts the right intelligence behind every task.
What it holdsOpen and frontier models, image and video models, evaluators, and the record of which model suits which task.
NeoCompute
runs what you build.
What it holdsHosting, pipelines, monitoring and support. Every request your live product serves passes through the router too.
You don't need to become a model expert
That's the point.
Can I override the router?
Yes. Every task shows the model the router chose. Pick a stronger or cheaper one yourself, and the quote updates to match.
Is a stronger model always better?
No. For bulk reading, summaries and templated pages a low-cost model does the same work at a fraction of the price. The premium model earns its place on hard reasoning, code across many files and final judgements.
Are the example prices live quotes?
No. The costs on this page come from a scripted demo. A real quote is produced by Neo for your job, before any work starts.
See every choice. Change any of them.
Bring an idea. Get a plan. See the price first.