Can you update
The app is great! Please update with the newest model. Thank you very much!
Great way to start, but feels unfinished.
This is a great, free starting point for someone who wants to play with AI chat, but doesn’t want every query being sent and stored in the cloud. The models offered have apparently been optimized for Apple silicon, and they do feel more responsive compared to other apps. The suggested Llama 3.2 works well, as does its bigger brother (if your phone has RAM for it).
My single biggest gripe is the apparent lack of ability to give the models standard instructions, like a “system prompt.” That omission is enough to knock this to four stars for me. It’s not a major issue for the default Llama model, or the bigger Llama model. But the Qwen models can be VERY verbose. I don’t want to have to keep telling the model “be brief.”
There’s also some fit-and-finish issues that aren’t show stoppers, but could use attention.
First, the initial model download stalled around 60% of the way in. There’s no obvious way to abort and retry. I killed the app and tried again. Fortunately that had no ill side effects.
Second, once a model is installed, I didn’t find a good way to delete it. I wound up deleting the app (purging my chat history) and reinstalling to remove the extra model.
Lastly, the number of models are limited. I realize that’s probably because of the Apple silicon optimizations. In the fullness of time, it would be nice to have some more options, even if they were paid upgrades.
Thanks to the developer for the effort. Again, it’s a great start; I’m picky. They clearly know what they’re doing, and I’m sure they’ll get there eventually.
My single biggest gripe is the apparent lack of ability to give the models standard instructions, like a “system prompt.” That omission is enough to knock this to four stars for me. It’s not a major issue for the default Llama model, or the bigger Llama model. But the Qwen models can be VERY verbose. I don’t want to have to keep telling the model “be brief.”
There’s also some fit-and-finish issues that aren’t show stoppers, but could use attention.
First, the initial model download stalled around 60% of the way in. There’s no obvious way to abort and retry. I killed the app and tried again. Fortunately that had no ill side effects.
Second, once a model is installed, I didn’t find a good way to delete it. I wound up deleting the app (purging my chat history) and reinstalling to remove the extra model.
Lastly, the number of models are limited. I realize that’s probably because of the Apple silicon optimizations. In the fullness of time, it would be nice to have some more options, even if they were paid upgrades.
Thanks to the developer for the effort. Again, it’s a great start; I’m picky. They clearly know what they’re doing, and I’m sure they’ll get there eventually.
Show more
🚨 False advertising: does not work offline
This app does not work offline. I have switched to Apollo for running offline LLMs. That one actually does work.
To reproduce: 1. Download a model and let complete.
2. Exit app.
3. Turn on airplane mode, turn off cell data and WiFi.
4. Open app and type to chat.
5. Chat returns a timeout error, long after the time interval it would have taken to start a response had the app been connected.
To reproduce: 1. Download a model and let complete.
2. Exit app.
3. Turn on airplane mode, turn off cell data and WiFi.
4. Open app and type to chat.
5. Chat returns a timeout error, long after the time interval it would have taken to start a response had the app been connected.
Show more
Works well, but not on flights.
This app is great. While it’s does work offline, the app requires a connection to hugging face, so it’s not always possible to use it during a flight.
For less powerful phones
My M2 iPad can handle bigger models, but Fullmoon’s in-app list is constrained to models they have packaged for mobile.
