Common questions
A model will not load
Large models can exceed the memory available on an iPhone or iPad. Unload the current model, close memory-heavy apps, and try the recommended smaller mobile model. Model size on disk is not the same as memory required while running.
A model download stopped
Reconnect to a stable network and retry. Confirm that the device has enough free storage for both the download and its working files. The model host can also be temporarily unavailable.
Does Axon send chats to Rapid42?
No. Normal supported MLX chat inference runs on your device, and Rapid42 does not operate a chat backend for Axon. Optional web search, model downloads, Apple voice dictation, and endpoints you configure have separate network behavior described in the privacy policy.
Why did web search contact the internet?
Web search is an optional network feature. When enabled and used, it sends the search query to configured or built-in search providers. Disable web search when you want an entirely local chat session.
How does Private Mac Link work?
Private Mac Link connects the app to an Axon endpoint you configure on your own Mac or private network. Axon rejects ordinary public-host targets for normal chat routing. Do not expose a local endpoint publicly without understanding the security implications.
What should I include in a support request?
Include the Axon version, iPhone or iPad model, iOS or iPadOS version, model name, and the exact visible error. Do not send private conversations, model files, API keys, passwords, or other secrets.