Tip Of AI
Discover AI tools
Llama is a lightweight desktop application created by the llama.cpp team and Hugging Face for running open AI models directly on personal computers. It works on Mac and Windows and offers an offline experience with no required configuration, usage charges, or API keys. Users can chat with installed models through a browser based interface or connect coding agents, editors, scripts, and other applications through a local API. The application provides an OpenAI compatible API at localhost:9931/v1 and also supports Anthropic compatible endpoints, streaming, tool calling, structured output, and vision. It automatically recommends models and adjusts settings such as quantization, context size, image input, speculative decoding, and batch size according to available computer memory. Supported model examples include Gemma 4, GPT OSS, and Qwen 3.8. Llama has a small download of about 1 MB and an installed application size of about 4 MB. Models are stored once in the Hugging Face cache and can be shared with llama.cpp and other tools. Models load only when needed and unload after five minutes of inactivity. The software is free, open source, works offline, and includes command line installation options for Linux.
