LLM Farm

LLM Farm runs large language models offline on iOS and macOS. It's for people who want on-device AI chat or need to compare how different models perform on Apple hardware before choosing one for a project. The app is open source under the MIT license, with ggml and llama.cpp handling local inference.

Model support includes LLaMA, Gemma, Phi, Qwen, Deepseek and Mixtral MoE, alongside architectures such as RWKV and Mamba. For image-based conversations, it supports vision models including LLaVA, BakLLaVA, MobileVLM, Yi-VL and Moondream. This gives prospective users a choice of text and multimodal models within the same app.

Model settings templates help reuse preferred configurations, and context restoration lets chats resume with saved context. Sampling controls let users adjust how models generate responses, including temperature, Mirostat and grammar constraints. These controls also make the app useful for testing model behavior, beyond everyday chat.

LLM Farm includes retrieval-augmented generation (RAG) for using reference material in conversations. Apple Shortcuts support connects the app to automations on Apple devices. Model inference runs on the device and works without an internet connection.

The app supports iOS 16 or later and macOS 13 or later. It can use Metal GPU acceleration, though its Metal support doesn't work on Intel Macs. The README says the app is temporarily unavailable in TestFlight and the App Store; source builds remain documented.