# Model as a Service (MaaS)

> A cloud runs the model for you, and you pay for each use.

**Field:** Artificial Intelligence  
**Category:** Cloud AI Platforms  
**Also known as:** MaaS, Serverless Model API, On-Demand Inference

## Think of it like

It is like a taxi. You do not own the car. You pay for each ride.

## Definition

Model as a service means a cloud runs an AI model for you, and you pay only for what you use. You send a request to an API and get an answer back. The bill counts the tokens that go in and come out. You do not pick, start or look after any servers. Clouds also call it serverless or on-demand use. It is the cheapest way to start, but speed can dip when many people use the same model at once.

## Example

Jack's side project gets a few hundred requests a day. He uses a model as a service and pays a few dollars a month, with nothing to run himself.

## Why it matters

It lets anyone use a top model on day one, with no hardware and no upfront cost.

## Common mistake

Expecting the same speed at every hour. Shared capacity can slow down or limit you when demand is high.

## Three principles

- **The cloud runs it.** No servers for you.
- **Pay per token.** In and out.
- **Shared capacity.** Speed can vary.

## Related terms

- [Serverless Computing](https://learn.benibix.com/term/infrastructure/serverless-computing)
- [Token Pricing](https://learn.benibix.com/term/development/token-pricing)
- [LLM API](https://learn.benibix.com/term/development/llm-api)
- [Provisioned Throughput](https://learn.benibix.com/term/ai/provisioned-throughput)
- [Inference Provider](https://learn.benibix.com/term/ai/inference-provider)
- [Tokens per Minute (TPM)](https://learn.benibix.com/term/ai/tokens-per-minute)
- [Serverless GPU](https://learn.benibix.com/term/ai/serverless-gpu)

## Told apart from

- [Model as a Service vs Provisioned Throughput](https://learn.benibix.com/compare/model-as-a-service-vs-provisioned-throughput): People often think provisioned throughput is just a faster, pricier kind of paying per token. It is a different deal. With model as a service, you share space and pay for each token. With provisioned throughput, you rent a fixed slice. You pay the same whether you use it or not, and get steady speed. Light or spiky use is cheaper per token. Heavy, steady use can be cheaper to reserve.
- [Serverless GPU vs Model as a Service](https://learn.benibix.com/compare/serverless-gpu-vs-model-as-a-service): People think a serverless GPU is the same as model as a service, and it is not. With model as a service, you call a model the provider runs and pay per token. With a serverless GPU, you bring your own model, and the service runs it on GPUs that switch on when called. You pay for the busy seconds. One rents a model. The other rents the chips.

![An app sends a request to a model the cloud runs. The bill counts tokens in and tokens out, with no servers to manage.](https://learn.benibix.com/diagrams/ca-model-as-a-service.svg)

---

Source: https://learn.benibix.com/term/ai/model-as-a-service
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-19
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
