FAQ
Agent workspace, explained clearly.
Answers for search engines, agents, and people evaluating Think before installing it.
What is Think?
Think is a local-first AI agent workspace for Apple devices. It is designed for local MLX workflows, files, tools, open models, and optional providers users choose.
Does Think require internet access?
Think is local-first and offline-capable after the app and required model assets are available on the device. Some setup, downloads, App Store access, remote providers, community links, or email support may require internet access.
Which devices are supported?
The website markets Think for iPhone, iPad, Mac, and Apple Vision Pro, using iOS, iPadOS, macOS, and visionOS.
Which local LLM models are relevant for iPhone?
Common small-model families for iPhone local LLM workflows include Llama 3.2 1B and 3B, Qwen3 0.6B, Qwen3 1.7B, Qwen3 4B, and Gemma 3 1B. Current Think availability should be checked in the in-app model catalog.
See the best local LLMs for iPhone guide and the Llama, Qwen, and Gemma guide.
How does Think handle privacy?
The product emphasizes on-device AI and clear provider choice. The core privacy claim is that prompts, files, and tool results can stay on the user's device for supported local workflows instead of being sent to a hosted AI service.
Does Think support open-source models?
The landing page says Think supports open-source AI models and Apple-native local execution where available.
How can I contact the team?
Email contact@tlon.sh or join the Discord community.