Guide · Mobile apps

AI in mobile apps: on the device or in the cloud?

The AI in a mobile app can run on the phone itself (fast, private and with no per-use cost, but with small models) or in the cloud (far more powerful and easy to update, but with latency, network dependence and per-use cost). The most common approach is to combine both: fast and sensitive tasks on the device, complex reasoning in the cloud.

Request a quote See how to decide

By Oliver Guirado · Updated on October 28, 2026 · 4 min read

The difference in one sentence

On the device, the model travels inside the app and runs on the phone’s processor: it needs no internet and the data stays put. In the cloud, the app sends the request to a server where a much larger model runs and returns the answer. Each option solves one type of problem well, and the decision affects your app’s design, cost and privacy. For the full product picture, see mobile apps with artificial intelligence.

How to decide: criteria

Criterion	On the device	In the cloud
Latency	Very low and stable; works offline	Depends on the network and the server
Privacy	Data may never leave the phone	Data has to be sent and agreements signed with the provider
Quality and power	Small models, for specific tasks	Large models, reasoning and conversation
Cost per use	No cost per request	Pay per use that grows with users
App size and battery	The model takes space and uses battery	Light on the phone, uses data
Updates	With the app or by downloading the model	Immediate, without touching the app
Compatibility	Depends on each device’s hardware	The same for everyone

Which tasks fit where

Better on the device

Image recognition and document scanning, object detection, dictation and simple commands, filters, simple classification and functions that must work offline.

Better in the cloud

Conversational assistants, long summaries, reasoning over documents, advanced semantic search and anything that needs large models or server data.

AI integration →

Hybrid

Pre-process on the phone (for example extract text from a photo) and reason in the cloud with only what is needed, reducing cost and data sent.

On iOS and Android there are system tools and open-source libraries to run models on the device (Core ML on iOS; LiteRT, formerly TensorFlow Lite, and ML Kit on Android). The specific choice depends on the case and the OS versions you want to support.

Cost, the most underestimated factor

In the cloud you pay per use: every conversation or image processed has a cost that multiplies by your active users.
On the device the cost is in development: optimising and maintaining models for many different phones means more up-front work.
Reduce what you send: filter, compress and summarise on the phone before calling the cloud.
Set limits and caching: per-user caps and reuse of answers where possible.

To estimate with numbers, use the chatbot and agent cost guide as a reference for what using language models implies.

Privacy and app store rules

Tell users what data is used and why, and ask only for the permissions needed (camera, microphone, location).
GDPR: if you send personal data to the cloud, you need a legal basis, minimisation and agreements with the provider.
Apple and Google: declare the use of data and AI in their app listings; keep them up to date.
Transparency: indicate when the user is interacting with an AI system and offer a human alternative.

How we decide it in a project

Define the AI features and what experience they need (real time, offline, conversation).
Classify each feature by latency, privacy, required quality and expected cost.
Test with prototypes on real devices, including mid- and low-range ones.
Design behaviour without a network (what can be done and what is postponed).
Measure in production consumption, errors, cost and satisfaction, and readjust the split.

If you have a specific idea, tell us about it from custom AI app development.

Frequently asked questions

Can an app’s AI work without internet?

Yes, if the model runs on the device itself. It is common for image recognition, simple dictation or classification. Large conversational assistants usually need the cloud.

Is on-device AI more private?

Generally yes, because data may not leave the phone. Even so you must inform the user and comply with the GDPR for any personal data processed.

Which is cheaper, on-device or cloud AI?

On the device there is no cost per request but more development work; in the cloud development can be quicker but you pay per use, which grows with users. It depends on volume.

Can the two be combined?

Yes, and it is the most common approach: fast, private or offline tasks on the phone and complex reasoning in the cloud.

Does it affect battery and app size?

Yes. A local model takes space and uses battery, so it is optimised and measured on real devices, including mid-range ones.

Shall we talk about your project?

Tell us about your case and we will reply within 24-48 hours with a proposed scope and an indicative quote.

Request a quote

Tell us about your project

Describe what you want to achieve and which systems you work with. We reply within 24-48 hours.

Name *Email *

Phone *What do you need? *

What is your approximate budget for this project? *

Project description *At least: number of sections, a website/app you like as an example, and features.

I have read and accept the Privacy Policy and Legal Notice and the Conditions. *

Data controller: App Design Consulting S.L., Tax ID: B19837509. Purpose: handling the sending of information and commercial prospecting related to our services and/or products. Legal basis: consent of the data subject. Rights: you may exercise your rights of access, rectification, restriction of processing, erasure, portability and objection regarding your personal data, as well as withdraw the consent given. Additional information: see the detailed Data Protection information on this page.

Fields marked with * are required

Artificial intelligence services and guides

Everything we do with AI for business, from the diagnostic to the product in production. Artificial intelligence for business.