Can You Run Ollama on a Phone? Android, iPhone and Options

Not officially. Ollama’s downloads cover macOS, Windows and Linux, and there’s no Ollama app for iPhone or Android. On Android you can still run it from the command line through Termux, which packages Ollama. On an iPhone you can’t run Ollama at all, so your choices are to reach an Ollama server on your own computer or to use an app that runs the same kind of model file directly on the phone.

Which route makes sense depends on whether you want Ollama specifically or just want a local model on your phone.

What are your options for Ollama on a phone? #

The first two keep you inside the Ollama world, with its model library and its API. The third drops Ollama but keeps what most people wanted from it: an open model, running privately, with no account and no cloud.

How do you run Ollama on Android with Termux? #

Termux is a Linux terminal environment for Android, and the Termux package repository includes an ollama package. That means you don’t compile anything yourself.

1. Install Termux. Termux’s developers point people to F-Droid or their GitHub releases for current builds, so check their site before grabbing an old copy.
2. Update the packages: pkg update && pkg upgrade
3. Install Ollama: pkg install ollama
4. Start the server: ollama serve & (the & keeps it running in the background of the session)
5. Pull and run a small model: ollama run <model-name>, choosing a model from Ollama’s library that’s well under your phone’s RAM. On an 8 GB phone, stay around 3 to 4 GB of model file.
6. Keep Android from killing it. Run termux-wake-lock or pull down the Termux notification and acquire a wake lock. Android is aggressive about stopping background processes, and a killed server looks like a crash.

Expect it to be slower than on a laptop. Ollama in Termux generally runs on the CPU, while phone apps built on llama.cpp can use the GPU (Metal on iPhone, OpenCL on some Snapdragon chips). The CPU works, but a 4B model that streams comfortably on the GPU can crawl on the CPU and heat the phone up quickly.

Termux is a good fit if you already live in a terminal, want the Ollama API on the phone for scripts, or want to test the same model you run on your desktop. It’s the wrong choice if you just want to chat with an AI on the bus.

Can an iPhone use Ollama? #

iOS doesn’t allow apps to run background servers like Ollama, and there’s no Termux equivalent on the App Store. What you can do is point an iPhone at Ollama running on your Mac or PC.

1. Make Ollama listen on your network. By default it only answers on the computer itself. Set the OLLAMA_HOST environment variable to 0.0.0.0 and restart Ollama so it accepts connections from other devices.
2. Find the computer’s local IP address and note Ollama’s default port, 11434.
3. Install a client app on the phone that supports Ollama servers, and enter http://<computer-ip>:11434.
4. Keep it private. Only do this on a network you trust. Don’t forward port 11434 on your router to the open internet. If you want access away from home, a private VPN between your own devices is the safer route.

This setup can run much larger models than any phone, because your computer’s GPU does the work. The catch is that it stops working the moment you leave home, the computer sleeps, or the Wi-Fi drops. It’s local to your house, not local to your phone.

Can you use your Ollama models on your phone directly? #

Mostly, yes. Ollama and phone apps built on llama.cpp both run quantized open models, and Ollama can import GGUF files, the single-file format most phone apps use. The simplest path is usually not to copy files out of Ollama’s own storage, which uses its own naming, but to download the GGUF build of the same model from Hugging Face.

Two rules keep this painless:

• Pick a phone-sized quantization. A Q4_K_M build is the usual balance of size and quality. Our GGUF quantization guide explains the names.
• Check RAM before you download. A model file needs roughly its own size in free memory, plus room for the conversation. See how much RAM you need to run an LLM on a phone.

The simpler route: run the model natively on the phone #

If what you liked about Ollama was private, offline chat with open models, a native app gets you there with less work and better speed.

Personal LLM runs open models through llama.cpp on the phone itself, using Metal on iPhone and OpenCL on Snapdragon phones with Adreno 700-series or newer graphics, with a CPU fallback elsewhere. It has a built-in catalog (Qwen 3.5 in 0.8B, 4B and 9B, Gemma 4 E2B and E4B, Ministral 3 3B and GLM 4.6V Flash), and every card shows a “Fits your device” badge read from your phone’s actual RAM. If the model you ran in Ollama isn’t in the catalog, you can add any GGUF by URL, including Hugging Face links, with an optional vision projector for image input. The steps are in how to add a custom GGUF model to your phone.

After the one-time download it works in airplane mode, there’s no account, and your chats stay on the phone. The free version shows ads when you’re online; a one-time purchase removes them.

Which should you choose? #

• You want the Ollama API on your phone for scripts or other apps: Termux on Android.
• You want big models and mostly use your phone at home: Ollama on your computer, with a phone client on the same network.
• You want a private AI you can use anywhere, including on a plane: a native on-device app. Our guides for Android and iPhone walk through the setup.

Plenty of people use two of these: a big model on the desktop for heavy work, and a 4B model on the phone for everything else.

Frequently asked questions #

Is there an Ollama app for iPhone? #

No. Ollama offers downloads for macOS, Windows and Linux only. An iPhone can connect to Ollama running on your computer through a third-party client, or run open models on the phone itself through an app that includes its own inference engine.

Can Ollama run on Android? #

Yes, through Termux, whose package repository includes Ollama. It runs from the command line and usually uses the CPU, so small models work best. You’ll need a wake lock to stop Android from closing it in the background.

Is running Ollama on a phone private? #

Running it inside Termux is as private as any local model: the prompts never leave the phone. Connecting to Ollama on your computer is private too, as long as the traffic stays on your own network and you don’t expose the port to the internet.

What’s the easiest way to run an Ollama-style model on my phone? #

Download the GGUF version of the model from Hugging Face and load it in an on-device chat app that accepts custom GGUF files. That skips the terminal and usually runs faster because the app can use the phone’s GPU.