Local LLM runner vs PocketLLM

Run LLM on your iPhone, 100% local, no network (Network needed once to download a model) Models: • Qwen2.5 0.5B Instruct — ~0.4 GB (4 GB RAM) • Llama 3.2 1B Instruct — ~0.8 GB (4 GB RAM) • Qwen2.5 1.5B Instruct — ~0.9 GB (4 GB RAM) • Llama 3.2 3B Instruct — ~1.9 GB (6 GB RAM) • Qwen3 4B Instruct 2507 — ~2.5 GB (6 GB RAM) • Qwen2.5 7B Instruct — ~4.4 GB (8 GB RAM) Also: • Add your own GGUF (for example, from Hugging Face) • Auto-estimated max context from GPU memory; you can change it Requires iOS 18 or later. Tested on iPhone 11, 14 Pro, 15 Pro with iOS 26. Qwen models © Alibaba Cloud. Licensed under Apache License 2.0. Llama 3.2 is licensed under the Llama 3.2 Community License, Copyright © Meta Platforms, Inc. All Rights Reserved. Built with Llama. On-device inference uses llama.cpp (MIT License).

Your AI runs on your phone. Not on someone else's computer. Download an open-source language model once, then chat with it anywhere — on a plane, in a tunnel, in a country with no signal. Nothing you type ever leaves the device. There is no account, no server, no upload, and no way for anyone to read your conversations, because there is nothing to read. TALK TO IT, OUT LOUD Voice conversation works the way a conversation should: you speak, it thinks, it answers aloud, and it starts listening again. Speak Cantonese, Mandarin, English, Japanese — it replies in the language you used, in a natural voice. Tap while it is talking to interrupt, like you would with a person. SHOW IT A PICTURE Load a vision-capable model and attach a photo. Ask what is in it, ask it to read the text in a sign, ask what is wrong with the diagram. CHOOSE YOUR OWN MODEL A curated catalogue of open models, sorted by size so you can see at a glance what your phone can actually run: Qwen2.5, Llama 3.2, Gemma 2, Phi-3.5, SmolLM2, DeepSeek-R1-Distill, and Qwen2.5-Coder for programming questions. One tap downloads; a colour tells you green, amber, or red for your device's memory before you commit. Already have a GGUF file? Import it. Any model in that format loads, including ones that are not in our catalogue. BUILT TO NOT WASTE YOUR SPACE See exactly how much storage models are using and how much you have left. Delete any model in one tap. Unload the running model to free memory without deleting the file. If a model is too large for your device, the app tells you before the download starts instead of letting you waste a gigabyte finding out. WHAT IT COSTS Text chat is unlimited and free, forever. Voice conversation and image understanding are free five times a day, plus one more each time you watch a short video. A single purchase removes ads and makes everything unlimited. No subscription. Requires roughly 1–4 GB of free storage per model. Larger models need more memory and a recent device.