SaaS Metrics
App Store
Locally AI by LM Studio
Run AI models like Llama, Gemma, Qwen, and more on your iPhone and iPad. Offline. Private. No login. No data collection. Powered by Apple MLX.
Meet Locally AI by LM Studio, the AI app that respects your privacy. No login. No data collection. Just pure intelligence, running entirely on your iPhone and iPad.
Why Locally AI?
- 100% Offline: No internet needed. Chat anywhere.
- Total Privacy: No login, no data collection. Your data stays on your device.
- Optimized for Apple Silicon: Powered by Apple MLX for lightning-fast responses.
- Multiple models supported: Run Llama 3.2, Google Gemma 2 & 3, Qwen 3, DeepSeek, and more.
- Chat, Analyze, Create: Ask complex questions, analyze images, or generate text.
Download Locally AI now!
Privacy Policy: https://locallyai.app/privacy-policy.html
Terms & Conditions: https://locallyai.app/terms.html
View Raw Thread
Developer & User Discourse
Escapable Tech • Aug 26, 2026
★ 5
This would be perfect, if I could use my lm studio on pc to do the heavy lifting, or if the free nicer models weren't locked behind a subscription.
Shadow Ratt • Aug 26, 2026
The kv cache seems to fill up fast on the 4B models to get more than a few messages. otherwise, it’s nice
somerandomdude... • Aug 22, 2026
Few issues I see with this version (since there's no feedback form):
- cannot swipe right when side scrolling in a code block; it opens the side bar
- scrolling us very jittery when a response is being generated
- cannot paste images from clipboard
Overall a really cool project and has promise
- cannot swipe right when side scrolling in a code block; it opens the side bar
- scrolling us very jittery when a response is being generated
- cannot paste images from clipboard
Overall a really cool project and has promise
VicenteSilvestre • Aug 21, 2026
Can attach PDF files. But model does not see them. I attach a PDF file and ask the model to summarize it. The model says that it cannot see any file or content. Large VL model running in LM Studio via LM Link works well otherwise.
LeighPDX • Aug 17, 2026
Doesn’t make any sense that the voice is processed in Apple’s cloud.
I-love-games-13 • Aug 16, 2026
Will save your life to have AI help anywhere.
Vyrtedzs • Aug 12, 2026
It’s ok
Cybernettr • Aug 10, 2026
The instructions sound simple:
1. Install LM studio on your PC.
2. Sign into your LM Studio account on both devices.
3. Turn on LM link.
On the phone app, in the Link to LM studio screen, there is a try it now button. I click it, and it sends an authorization link to my email. I click the link, and the device/app on my PC should be authorized. However , am caught in a loop. When I go back to the app, it still wants me to click the the button so it can send me the authorization once again. There is no button that appears that says the phone and device are now authorized and ready to use. i've been trying to figure this out for more than an hour. This appears to be more trouble than it's worth.
1. Install LM studio on your PC.
2. Sign into your LM Studio account on both devices.
3. Turn on LM link.
On the phone app, in the Link to LM studio screen, there is a try it now button. I click it, and it sends an authorization link to my email. I click the link, and the device/app on my PC should be authorized. However , am caught in a loop. When I go back to the app, it still wants me to click the the button so it can send me the authorization once again. There is no button that appears that says the phone and device are now authorized and ready to use. i've been trying to figure this out for more than an hour. This appears to be more trouble than it's worth.
BetterN'Better • Aug 5, 2026
Could you please add a setting to specify a custom OpenAI-compatible base URL and API token? I’m running LM Studio on my Mac Studio, and I’d like to point the app there directly, as the current LM Link integration has been unreliable for me.
steve_yoo • Jul 28, 2026
Works great aside from few display glitches and inability to display equations properly with Qwen 3.5/4B model. Workaround is to paste output into Obsidian that can display LaTex output.
E.g. Prompt: Provide more info on Electric field mill: A grounded, multi-vaned metal shutter shaped like a propeller. It is mounted directly above the stator on a spinning motor shaft.
E.g. Prompt: Provide more info on Electric field mill: A grounded, multi-vaned metal shutter shaped like a propeller. It is mounted directly above the stator on a spinning motor shaft.
Sam171717 • Jul 25, 2026
On my MacBook Pro, the text are really, really tiny. Using keys Command + does NOT increase the font size. In settings, there is not ption to change fonts size - which is a very, very simple thing to implement.
Nevertheless, if you can’ see it, then you can’t use it. Now I have to find another AI to do the job this one can’t.
Nevertheless, if you can’ see it, then you can’t use it. Now I have to find another AI to do the job this one can’t.
TomWilson • Jul 23, 2026
Does exactly what it says on the tin, no bugs or errors that I’ve come across, there are some glaring omissions in the UI though.
The most prominent one is there’s no way to regenerate or edit AI responses in the app even though that’s built into the (amazing) LM Studio, so obviously they can do it, it’s just a bizarre choice not to be able to do that remotely. Update that and I’ll come back and change this to five stars because the software works very well, but that’s IMPORTANT functionality that has just been omitted for whatever reason
The most prominent one is there’s no way to regenerate or edit AI responses in the app even though that’s built into the (amazing) LM Studio, so obviously they can do it, it’s just a bizarre choice not to be able to do that remotely. Update that and I’ll come back and change this to five stars because the software works very well, but that’s IMPORTANT functionality that has just been omitted for whatever reason
LikeARyanStoneCowboy • Jul 22, 2026
Absolutely brilliant app! Use the lm link to use a more powerful machine thru your phone
Bob Greeny • Jul 22, 2026
Please make it so iOS app allows iOS shortcuts to use a model thats on another device. Currently the dropdown only shows local models. This is the only value I would have from the app and it’s not supported and I think it’s a bug or dev’s forgot to do it? Please fix ASAP!!!
stawberrykiwi • Jul 22, 2026
All equations don’t render, instead showing up as a blank space making it impossible to ask any math-related questions. Please fix. Or even adding a plain text view that can be toggled on by the user to display the output in plain markdown would be better.
Another annoyance is that colored words in code blocks flash between red and white while generating causing an eyesore (and hopefully can be fixed for the safety of epilepsy-prone individuals).
I would give five stars otherwise.
Another annoyance is that colored words in code blocks flash between red and white while generating causing an eyesore (and hopefully can be fixed for the safety of epilepsy-prone individuals).
I would give five stars otherwise.
chaotux • Jul 21, 2026
Nothing I’ve tried will link the LM Link and allow the quoted feature. No support anywhere to get answers as to why it’s not connecting.
Petrenkas • Jul 21, 2026
In the iOS version of the Locally app, mathematical formulas are not rendered correctly.
Instead of displaying formatted equations, the app shows the raw LaTeX source code. When I copy this LaTeX code and paste it into any external editor that supports LaTeX/MathML or Markdown math, the formulas render correctly, which suggests the issue is in the client-side rendering, not in the generated content.
Instead of displaying formatted equations, the app shows the raw LaTeX source code. When I copy this LaTeX code and paste it into any external editor that supports LaTeX/MathML or Markdown math, the formulas render correctly, which suggests the issue is in the client-side rendering, not in the generated content.
Anonymous Coward9 • Jul 19, 2026
The app is great and does what it promises. However, there are some “quality of life” things I’ve come to take for granted from LM Studio that aren’t currently supported here, such as editing prompts, responses, or thoughts. Likewise, you can only have one personalization prompt. All the same, it’s cool that everything is so mindful of privacy.
twisted_nematic57 • Jul 18, 2026
It technically does work and runs LLMs on your iPhone. A major gripe I have with it though is that LaTeX expressions are FREQUENTLY displayed incorrectly. That is, the model is writing valid LaTeX but Locally’s renderer doesn’t display it properly. Before the latest update it was shaky—long expressions couldn’t be scrolled at all, and some shorter ones had weird layout issues and looked like they were off-place a couple dozen pixels—but after the latest update, all LaTeX fields just appear blank entirely. I would’ve given this app 5 stars if it weren’t for LaTeX issues.
Also, I highly recommend adding some kind of web search option like Ollama Web Search. That’d be mighty useful for letting the model know about current events and info it may not be able to come up with on its own.
Also, I highly recommend adding some kind of web search option like Ollama Web Search. That’d be mighty useful for letting the model know about current events and info it may not be able to come up with on its own.
Alexphotomaker • Jul 17, 2026
IPhone 17 pro max
Bonsai 27B - doesn't work after 2nd message.
Gemma 4 E2B - doesn't support at all.
Bonsai 27B - doesn't work after 2nd message.
Gemma 4 E2B - doesn't support at all.