fake news: no LFM3 available
"""
What's New ›
Version 2.3.0
14h ago
Apollo 2.3 brings LFM3, Liquid Al's newest generation of on-device models. LFM3 packs more intelligence into less space, so it runs comfortably and privately right on your device.
"""
Nope. This was the whole reason I reinstalled the app (it was offloaded). I thoroughly tried all sorts of things, including eventually resetting the app via your in-app option to do so (btw, it does not seem to restart the app like it says it does), and then finally fully deleted the app anp and its data. The only difference: it shows less models at the intro model selection screen, but still no LFM3 models.
iPhone 17 Pro Max, iOS 27.[latest available] (27.2, I believe).
Back to the excellent NoemaAI app (and the less advanced Off Grid app, and Pocket Pal app), I suppose (although Noema's MLX support rejected your model, not recognizing the label for the new lfm3 family or model, and no CoreAI models with image/multimodality support were found, IIRC).
Replies are very welcome.
What's New ›
Version 2.3.0
14h ago
Apollo 2.3 brings LFM3, Liquid Al's newest generation of on-device models. LFM3 packs more intelligence into less space, so it runs comfortably and privately right on your device.
"""
Nope. This was the whole reason I reinstalled the app (it was offloaded). I thoroughly tried all sorts of things, including eventually resetting the app via your in-app option to do so (btw, it does not seem to restart the app like it says it does), and then finally fully deleted the app anp and its data. The only difference: it shows less models at the intro model selection screen, but still no LFM3 models.
iPhone 17 Pro Max, iOS 27.[latest available] (27.2, I believe).
Back to the excellent NoemaAI app (and the less advanced Off Grid app, and Pocket Pal app), I suppose (although Noema's MLX support rejected your model, not recognizing the label for the new lfm3 family or model, and no CoreAI models with image/multimodality support were found, IIRC).
Replies are very welcome.
Show more
Uusayerong
Only if
This is excellent, I wish you could update it to allow selection of Vision capability and Audio capability for models provide by custom servers.
I run oMLX, it supports vision, video, audio (STT, STS, TTS), all of which are available but cannot be utilised because the app itself isn't aware of it.
My local server models come with 1,048,576 context window, meaning it could easily support a local memory saved on the device to send with any chats but unfortunately the app does not support it.
At the very least being able to use vision for image and video with a local server would be great.
Also, the ability to reduce the size of icons and fonts; it feels too zoomed in.
Thank you for the great application.
I run oMLX, it supports vision, video, audio (STT, STS, TTS), all of which are available but cannot be utilised because the app itself isn't aware of it.
My local server models come with 1,048,576 context window, meaning it could easily support a local memory saved on the device to send with any chats but unfortunately the app does not support it.
At the very least being able to use vision for image and video with a local server would be great.
Also, the ability to reduce the size of icons and fonts; it feels too zoomed in.
Thank you for the great application.
Show more
PPrivacy_for_life
Some issues
Nice app, but it keeps changing the model on openrouter and it can’t render latex outputs for some reason. Maybe it’s the model I’m using but it should work since it works on open routers test workspace.
Pph9214
Cannot remember the last model
Very frustrating that I have to tap 6 times to pick the model I want to use. Every single time.
RCRobert Chitoiu
Doesn’t work for local serving
The AI keeps cutting off context length based on the internal client side context limit and not the server side limits. Please fix.
FFFunctionality fanatic.





