The iOS app is very simple right now. That is good from a ease of use perspective, but it does not say what model is used and there is no way to choose a different model. In some quick testing, the responses are fast but extremely inaccurate. Most of the responses were completely unrelated to the prompt. The best responses were only partially correct, but were at least dealing with the right information.
I am very happy to see more development focused on local AI that does not require subscriptions or sharing personal data and are able to work completely offline. Unfortunately, this app is not useful currently. I am assuming they chose to use a very small model to support the most devices and provide fast responses, but the output is extremely inaccurate. I will not be using this app until they support a better model like Llama 3.2. PocketPal is still the best local llm app I have found on iOS and Alpaca is great on Linux.
I am very happy to see more development focused on local AI that does not require subscriptions or sharing personal data and are able to work completely offline. Unfortunately, this app is not useful currently. I am assuming they chose to use a very small model to support the most devices and provide fast responses, but the output is extremely inaccurate. I will not be using this app until they support a better model like Llama 3.2. PocketPal is still the best local llm app I have found on iOS and Alpaca is great on Linux.