Mobile Offline LLMs: What do you use them for?

A developer has been experimenting with open-source MLX and GGUF models on iPhone hardware for about a year, using them for various tasks. These include building tools for web search, URL scraping, summarization, research, analyzing local media and documents, and providing on-the-go coding assistance. Despite context window limitations of 8-16k tokens, continuous compaction allows for virtually unlimited exchange chats. The developer notes that while many focus on laptop/desktop LLMs, mobile applications remain largely unexplored and seeks to understand how others utilize mobile models.