How to Run a Local Master’s in Applied Management Course on Your Phone (and Why You Might Want To)
There are now enough freely available large language models (LLMs) that running them locally on your computer has become commonplace, and in some ways they’re even better – they’re more private because you don’t have to send anything to the cloud, and they run offline.
Running these advanced AI models on phones hasn’t been entirely straightforward (that’s why Siri AI’s most advanced features require the latest iPhones), but we’ve now reached the point where most phones are powerful enough, and some models are compact and efficient enough, that this is becoming a reality.
The benefits are the same as on a desktop computer—you get an LLM that’s always available and private, and nothing is sent back to Google, OpenAI, Anthropic, or anyone else. The downsides are that your AI performance will be slower and more limited, as you’ll be working with smaller, less powerful models, and your battery life may be reduced (those AI chats can be quite resource-intensive).
Even with these tradeoffs, the local AI you can run on your phone will still be more than capable of handling everyday tasks and chatting—meaning you can give Gemini or Siri (and maybe your AI subscription) a rest.
Telephone number required for conducting the local master’s program.
Any iPhone or Android smartphone released in the last couple of years should be able to run a local LLM system quite well; the AI boom means manufacturers have started designing their devices specifically for this type of use. Older phones are also suitable, but you may need more compact models, and performance will be worse.
In fact, RAM capacity is more important than chipset performance when it comes to running an AI model. 6 GB or more is required for satisfactory performance, while 8 GB or more is better for larger models. Less than 8 GB should be used for models of 1-2 billion in size—the “b” stands for billion and denotes the number of parameters in the model (essentially, how intelligent and versatile it will be).
As for storage space, you don’t have to worry too much. Even the largest models, using 7-8 bytes of memory, take up around 5 GB, so you can even store multiple AI models on your phone and switch between them as needed.
If you buy a new flagship iPhone or Android smartphone today, it will come preinstalled with local AI models for quick tasks that can be completed without accessing the cloud. However, these models, created by Apple and Google, are not directly accessible to you, the user—they are used when Siri AI and Gemini deem them the best option.
Applications and models required for on-site delivery of the LLM program.
There are several apps that can help you install and run AI models on your phone. The most popular are Atomic Chat ( for Android or iOS ) and PocketPal AI ( for Android or iOS ). There are some differences—for example, Atomic Chat is easier to extend to work with desktop LLM models, while PocketPal AI is a bit more lightweight—but both are great for getting started with local AI on your phone.
As for the models themselves—sometimes called SLMs or small language models, as they’re designed to work in more constrained environments—you also have a wide range of choices. For example, Gemma is the general name for open-source models that Google provides for free, and some of them are specifically designed to work in more constrained environments with fewer resources (like your phone).
Meta has its own open-source AI models, called Llama, and Microsoft has Phi-4 models, which are also highly regarded for their efficiency. Just look for versions with the fewest parameters preceding the letter “B” (or with “mini” in the name) to find the packages that are best suited for your phone.
Currently, these image management systems (SLM) primarily work only with text, although some newer, larger, and more advanced models can analyze images and files. If you want to create images and videos, you’ll have to use traditional cloud-based AI models, at least until the next technological breakthrough.
Testing the local Master’s degree program in law in practice.
To evaluate the usefulness of one of these local AI models on a phone, I installed PocketPal AI and one of the more compact Google Gemma models on my Pixel 9 Pro—no longer a flagship phone, but not particularly old either. I liked the small selection wizard that launches in PocketPal AI and immediately directs you to the right AI model for your phone.
After downloading and installing a couple of models, getting them up and running was fairly easy. PocketPal AI also provides access to additional “friends” (hence the app’s name) who can tailor AI models to your needs—the standard Pip feature is a perfectly adequate replacement for what you might be used to in the default Gemini or Siri AI apps. Suggestions and follow-up actions work as usual, and chat history is saved by default.
When running LLM models on a phone, there’s a noticeable (and expected) slowdown in response time, as well as a noticeable difference in AI model sizes—choosing a smaller model will yield significantly faster responses, even if it’s not as intelligent or complete. It’s worth experimenting with several models to find the optimal balance between performance and speed.
In the absence of web search and up-to-date information, this method is best suited for brainstorming, analyzing and refining existing text, writing new text, and quickly obtaining facts or comparisons (“Name a movie similar to…”). As always, beware of hallucinations and don’t take the AI’s word for guaranteed accuracy.