Decent but not 5 stars
I’ve gotten all functions too work on this app EXCEPT importing llms from hugging face with my iPhone 16 plus running 8gb of ram I can’t load unsloths Gemma or qwen models only the built in one hope too see this fixed soon as the built in model tends to have looping issues when asked complex questions like “name every iPhone model ever made with a plus variant”
Show more
Response from developer
Thank you for the feedback! I regularly run the Q4_KM versions of both gemma3 4b and qwen3 4b on my iphone 14pro without any issues. Can you please email me the models and quants you used so i can check and fix any issues?
Really impressive, surprisingly useful
This is a great app. The combination of the astonishing Qwen3 1.7b model (built-in), the capacity to search the web, and to think, enables it to punch way above its weight. It’s just the right size to run quickly and comfortably on my m2 iPad. Very highly recommended if you have a modern ipad or iphone.
Show more
Awesome app - Keep going
Really what i was looking for in an on device LLM without tracking etc
5 stars BUT
Lovely app, by far the most useful for situations if online is needed as other apps don’t have this function, suggesting a voice chat mode in the future and file chat(dropping pdf etc) whether it be a pdf book of DnD rules or a .py of a snake game, being able to do all that here would be amazing
Response from developer
Hello, thanks for the review! Memories does work (check out the source on github) so its not a spoof, I am working on making it better. It just doesn’t work well right now. Thanks for your support!
5 stars BUT
I don’t want to lower the rep before the app gets to grow but it’s very clear to me the memories are not actually injected at all or therefore seems to be a spoof or not a full feature that’s working yet… that’s one thing I ask of is the memory actually being a possibility here, and maybe a voice chat mode in the future and that’s it pretty much(also pdf etc dropping files for multimodal)
Show more
Response from developer
Hello, thanks for the review! Memories does work (check out the source on github) so its not a spoof, I am working on making it better. It just doesn’t work well right now. Thanks for your support!
Wow!
This is a remarkable application developed by a talented developer, and it's also free and open source. At its core, the app combines the central functionality of applications like Perplexity and Perplexica into a single package that runs smoothly and locally on an iPhone. (How is this even possible I would have thought even a year ago…) The developer has managed to integrate SearXNG, a small Qwen3 model, and an embedding LLM into one easy-to-use package that simply works.
There are still a few glitches with the application. Sometimes the LLM mysteriously fails to load, which I think has more to do with iPhone memory management issues than anything else. But usually force quitting the application and re-launching fixes that problem.
But, overall, this app fills an important space by enabling lightning-fast local responses to user research queries. And all of this is done with privacy as a top priority.
Sincere thanks to the developer!
There are still a few glitches with the application. Sometimes the LLM mysteriously fails to load, which I think has more to do with iPhone memory management issues than anything else. But usually force quitting the application and re-launching fixes that problem.
But, overall, this app fills an important space by enabling lightning-fast local responses to user research queries. And all of this is done with privacy as a top priority.
Sincere thanks to the developer!
Show more
Outstanding
This app hits the sweet spot of privacy, online search, speed and function, with a great user interface to an amazingly capable local AI. Highly recommended.
This app fills an important niche
I can confidently say that this is the only mobile solution of its kind that I could recommend to friends and family. This app is exactly what I’ve been looking for. It works out of the box with a very capable model (Qwen3 1.7B) at speeds slightly faster than I read and doesn’t take long to fetch sources. No advanced configurations required or internet required if you need to chat in airplane mode as you’d expect.
I love the peace of mind that using local models with web search on my Mac with apps like Msty and browser extensions like Page Assist. But the mobile counterparts of other chat UIs (e.g. OpenWebUI, SillyTavern) are far to complicated for the average non-dev user like me to configure. MyDeviceAI feels a missing niche that I’ve reached for daily in the past while away from my MacBook.
Can’t wait to add more personalized info to the app’s memory for context (something Perplexity does not have as of writing) and for future updates. Hope this app sticks around for the long run and the dev gets the well-deserved the support necessary to do so!
I love the peace of mind that using local models with web search on my Mac with apps like Msty and browser extensions like Page Assist. But the mobile counterparts of other chat UIs (e.g. OpenWebUI, SillyTavern) are far to complicated for the average non-dev user like me to configure. MyDeviceAI feels a missing niche that I’ve reached for daily in the past while away from my MacBook.
Can’t wait to add more personalized info to the app’s memory for context (something Perplexity does not have as of writing) and for future updates. Hope this app sticks around for the long run and the dev gets the well-deserved the support necessary to do so!
Show more
Fantastic private app
Like your own private version of perplexity. Where have you been all my life. Amazing and worth real money. Thanks to the dev. Try it. I wish it worked in iPad but it’s amazing on iPhone.
