To run Gemma 4 on a phone, install an app with on-device model support, download the E2B version (about 3 GB), and chat offline. Choose E4B (about 5 GB) only if your phone has plenty of memory.
E2B or E4B?
| Version | GGUF size (Q4_K_M) | Best for |
|---|---|---|
| Gemma 4 E2B | About 3.1 GB | Most phones with 6 to 8 GB of RAM |
| Gemma 4 E4B | About 5.0 GB | Phones with 12 GB of RAM or more |
Sizes are from Unsloth’s GGUF builds on Hugging Face in November 2026. A LiteRT build of E2B is about 2.6 GB.
Set it up in opchat
- Install opchat for iPhone or Android.
- Open Settings → Models → On this device.
- Choose Gemma 4 E2B from the recommended list.
- Download it on Wi-Fi, then start a chat.
What Gemma 4 can do in opchat
Gemma 4 is one of the local models in opchat that can use tools. It can set reminders and save things to memory, and with Pro it can read and add to your calendar. Keep requests simple, since small models can get multi-step tasks wrong.
Other ways to run it include PocketPal AI and Google AI Edge Gallery.