Mollama

Downloads of MLX models in the app is extremely slow (even with high-speed internet), download progress isn’t accurate (it may still show download in progress, when the model is already downloaded), and once models are downloaded, they aren’t loaded when selected (system and app memory usage remains unaffected when selecting larger models). When a non-built-in model is selected, and the simple prompt "hi" is given, the app is busy, and returns nothing, except at some point flashing an orange "unsupported model" (or something like that" message. Things work with the built-in model, but I don’t need an MLX app to run something as trivial as a 0.5B micro model on an M4Max with 128GB RAM, when I could run that model on an old iPhone… If only the bugs were fixed, and this would be a great start.

The app is very accessible. Except when downloading a model. The app freezes the phone until it is finished downloading. The MLX model selection page is a great way to download models, but the model selection is very limited, and most phones won't run models that are two gig+, also only one uncensored model in the list and it's too big. Chat history is good, with auto named chats. The chat window however doesn't have the ability to add files only a camera button that doesn't work. Also when the chat gets too long, it refuses to send any more prompts Overall, it's a good app if you want something simple, Secure on device, and small basic models.

I just wanted to report that I downloaded a 4 gb random model and trying to talk with it on an iPhone 16 pro max makes the app crash immediately. Just figured I'd report that. I am running on a developer beta which probably matters, but I wasn't expecting a 4 gb model to crash the app. I'll update my review if I can get any other downloaded model working