Local LLM runner vs koke

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).

koke is a private AI assistant that stays useful when the internet does not. Download a model before you leave, then keep chatting on flights, underground, while traveling abroad, or anywhere with weak or no reception. Airplane Mode is no problem—AI inference runs entirely on your iPhone or iPad. Your prompts and responses stay on your device. No account, sign-in, cloud processing, or recurring subscription is required. WHY KOKE • Works offline after model download • Useful on flights, subways, and in dead zones • Fully on-device AI inference • Private conversations that stay on your device • Multiple optimized LFM2.5 models • Model downloads verified with SHA-256 • Native interface for iPhone and iPad • Background generation on supported devices KOKE PRO A one-time purchase unlocks all available models and custom GGUF model import. There is no recurring subscription. A network connection is required only when downloading a recommended model. Model files are downloaded from the official Liquid AI repository on Hugging Face. Prompts and generated responses are never included in model download requests.