Extensions

Multi-Model Routing Add-on

Simple tasks go to a cheaper, faster model, harder ones to a stronger model, inside the product you already run.

Sending every task to the most powerful AI model is like sending every parcel by express courier. This add-on lets the product you already run send simple tasks to a faster, cheaper model and harder ones to a stronger model, using routing rules you approve. A backup model takes over if a provider goes down, and all calls run from your own accounts.

Sends simple tasks to cheaper, faster models
Keeps harder tasks on a stronger model
Switches to a backup model if a provider fails

$2,500 10–20 working days

Choose delivery and add to cart Ask an expert

Wooden track splitting three ways at a brass switch, each task sent down the path to the right model.

What it is

Multi-Model Routing Add-on lets an existing Upmark Labs product send each task to one of several AI models. Without it, the product always uses the same model. Routing rules you approve decide which model handles what, with a fallback if one is unavailable. It is the single-product version of AI Model Router.

Why businesses buy it

AI usage costs grow with every conversation, and many of those tasks do not need the most expensive model. Routing them sensibly can lower your monthly usage bill and speed up replies, while keeping quality where it matters. It also protects you from a single provider outage. The written scope sets the rules per task type, the backup for each route and testing with the parent product, so savings never come at the cost of answers you cannot trust.

Problems this solves

Every task goes to one AI model, even simple ones

Multi-Model Routing lets the product send each task to one of several AI models instead of always the same one. Routing rules you approve decide which model handles what.

The product stops working when its only model is unavailable

A fallback is set so another model takes over if the usual one is unavailable. The fallback is tested together with the parent product before anything goes live.

You want model accounts and usage held in your own name

Model calls run from accounts in your name, not ours. You see the usage directly and keep the relationship with each provider, as the updated documentation describes.

Scope card SKU UPM-A0034

Multi-Model Routing Add-on

Everything this build includes, written down before you pay. Choose how you want it, own the code or license it yearly, then the delivery speed and any add-ons.

$2,500 one-time build price

What is included

Channel: As per the parent product
Integrations: As described in the add-on
Knowledge: As per the parent product
Workflows: Lets an existing product route tasks between several AI models
Languages: As per the parent product
Testing: Tested together with the parent product
Revisions: 1 round
Handoff: Written handover report
Support: 14 days after launch

A good fit if

Your product handles a mix of easy and hard tasks, and every one goes to the most expensive model today.

You want a backup model ready if your main provider slows down or goes offline.

You already hold accounts with more than one AI model provider, or are willing to open them.

You run an Upmark Labs product, or are buying one at the same time.

Not the right fit if

You want one gateway for all AI traffic across your company. AI Gateway is the wider build.

You want routing built as its own system rather than inside one product. AI Model Router does that.

You only want to know when spending climbs. Cost and Usage Alerts sends those warnings.

You want to compare models on quality before choosing. Evaluation Test Suite gives you the test set first.

How it works

Tasks are sorted. We list what your product does, such as short replies, long summaries or classification, and how demanding each one is.
Models are matched. Each task type is assigned a model from the providers you hold accounts with, and the choice is agreed with you.
Rules and fallback are built. The router picks the model per request, and if that model fails, the request moves to the agreed backup.
Quality is checked. The same sample tasks are run through the routed setup to confirm answers stay acceptable on the cheaper paths.
You review the results with one revision round, then receive updated documentation and a written walkthrough.

What you provide within 72 hours

Your portal lists these the moment you order. Delivery starts the working day after everything is in and your access is checked.

The order reference of the product getting routing
API access to each AI model provider you want used, in accounts in your name, shared securely as agreed at intake
Any provider you must avoid for policy or data reasons
Your priority between cost, speed and answer quality
One person to approve the routing rules

The 72-hour rule. Send everything on your checklist, including access, within 72 hours of ordering so your build starts on time.

Timeline

Start Task types and provider accounts confirmed

Week 1 Model assignment per task agreed

Week 2 Routing rules and fallback built

Week 2 Sample tasks run on each route

Week 2 Your review and one revision round

Go-live Routing live, walkthrough recorded

Delivery options, in working days (Monday to Friday) from completed intake. They can extend because of national or public holidays, technical issues outside our control (such as platform outages, third-party changes, account approvals or access delays), or natural disasters and other events beyond our control. If a delay happens, we tell you in writing with a new date.
Option	Delivery time	Price
Standard	10–20 working days	$2,500

What you receive at handover

Routing between several AI models added to your existing product
The routing rules you approved, including the fallback
Model calls running from accounts in your name
Updated documentation, a test report, a handoff note and a written walkthrough

Expected benefits and how to measure them

Can match each task to a model that suits it
Helps the product keep working during a model outage
Gives you more choice over model providers

Model spend: Compare monthly model bills from your provider accounts for the two months before and after.
Fallback events: Count times the fallback took over each month from the product's logs.
Output quality by task: Review a sample of outputs per task type monthly and rate them on the same scale.

Before you start: Record your current monthly model spend and any outages in the last quarter. Note which tasks matter most for quality.

These are things to track, not promised results. Outcomes depend on your data, your team and your market.

Running costs

Model usage is billed by each provider to your own accounts. Routing is usually set up to move simple tasks onto lower-cost models, but the actual bill depends on your volumes and the providers you choose, so we do not promise a figure.

Ask about running costs

Custom requirements

Routing across several products at once, self-hosted models, or routing decided by a trained classifier fall outside this add-on. Note them at checkout for a written Change Quote. Larger or bespoke projects go through the Custom AI request form instead.

Request a custom version

Keep it running

Model providers change prices and release new models often. Swapping a model on one route uses the change hours of the parent product's maintenance, which grows in line with the add-on's price.

Adds $125/mo to maintenance

Covered by the main product's maintenance, which grows by this amount.

How maintenance works

How this compares

	Do it yourself with a SaaS tool	Hire a custom agency	Upmark Labs
Price	A routing service often adds a fee per request	An agency scopes it as architecture work	Ours is one fixed price, shown before you buy
Accounts	Traffic may pass through their account	Depends on the brief	Ours calls models directly from accounts in your name
Fallback	Varies by plan	Set per project	Ours includes an agreed backup model per route

Questions about this product

How many AI models can the routing add-on use?

Several, from the providers you hold accounts with. The practical number is set by your task types: most products need one model for simple work, one for demanding work and a fallback. We agree the list at intake, and each model is tested on the routes it serves.

Will cheaper models make answers worse?

Not on the routes where they are tested and approved. We only send a task type to a lower-cost model after its sample answers pass the quality check you agree. If a cheaper model falls short for a task, that task stays on the stronger model.

What happens if a model provider goes down?

The request moves to the backup model agreed for that route. The customer still gets an answer, and the switch is recorded so you can see how often it happened. When the main provider recovers, traffic returns to it automatically.

Do I need accounts with several AI providers?

Yes, for the models you want used. Each account is in your name and billed to you, so you keep the relationship with every provider. We help you decide which ones are worth opening during intake, based on the tasks your product handles.

Can model routing be added without rebuilding our product?

Yes. Routing is added to the product as it runs today, with the rules you approve, and the existing prompts and knowledge stay as they are. It can also be ordered with a new product. You choose the delivery speed at checkout, and each speed has a fixed price and date shown before you pay.

Can I change which model handles a task later?

Yes. Each route is documented, so swapping a model is a contained change. Under the parent product's maintenance, a swap uses the monthly change hours. A new route for a new kind of task can be quoted separately if it needs more than maintenance covers.

How does the routing add-on differ from AI Model Router?

The add-on adds routing inside one existing Upmark Labs product, using that product's setup and task types. AI Model Router is a standalone build that routes requests as its own system, weighing cost, capability and speed across your applications. Choose the add-on for one product, and the router when several applications need routing.