Pocket: the simplest way to use private LLMs. It works fully offline and ensures your privacy when you don’t have an internet connection. Pocket runs models on-device, optimized for Apple Silicon to provide fast and efficient performance.
Run powerful AI models like Llama, Gemma, and Qwen directly on your iPhone and iPad. Completely offline. 100% private. No accounts, no logins, and zero data collection. Your data never leaves your device.
Personalize your experience: adjust the theme, fonts, and system prompt directly within the app to match your preferences.
With Pocket's shortcut integration, you can easily chat with your local model or get an output to use with other actions, all while keeping everything local and secure.
Free, local, private — your data stays on your device, always.
Pocket supports a wide range of models to cater to various needs, including the following language models:
Llama 3 2 1B 4bit
Llama 3 2 3B 4bit
Llama 3 8B 4bit
Mistral 7B Instruct 02
Mistral 7B v02 4bit
Phi 4bit
Gemma 2 2B IT 4bit
Gemma 2 9B IT 4bit
Gemma 2B Quantized
OpenELM 270M 4bit
Qwen 3
Deepseek
Whether you're working on text generation, language processing, or complex tasks, Pocket has a model for you.
In addition to language models, Pocket also supports image models for users who want to generate visual content. With just a few taps, users can switch between text and image models, creating everything from written outputs to images—all processed locally on your device.
Customization is key: you can easily choose a new model or switch between existing ones within the app. Whether you’re switching between models to experiment with different capabilities or looking for specific output types, Pocket gives you full control.
With the ability to generate images when desired, Pocket can assist in visual projects alongside text-based tasks. All of this is done securely and privately, ensuring no data is stored or shared outside the app.
You may purchase an auto-renewing subscription through an In-App Purchase.
• Auto-renewable subscription
• pocket plus
• 1 week ($6.99) duration
• 1 month ($14.99) duration
• 1 year ($99.99) duration
• Your subscription will be charged to your iTunes account at confirmation of purchase and will automatically renew (at the duration selected) unless auto-renew is turned off at least 24 hours before the end of the current period.
• Current subscription may not be cancelled during the active subscription period; however, you can manage your subscription and/or turn off auto-renewal by visiting your iTunes Account Settings after purchase
• Terms of use: https://www.apple.com/legal/internet-services/itunes/dev/stdeula/
• Privacy Terms: https://oneai.framer.website/privacy
Very nice app. I’m paying for premium. I love the research feature for learning about something thoroughly. That’s unique enough to make me pay for it. I do think there really needs to be a focus on improving the text to speech so that text can be read out loud with a better model. This only works sometimes to begin with/may be very slow, and the voice is pretty bad. Pls work on that and I’ll continue paying.
Developer Response
Thank you so much for your support and for being a premium subscriber! I am really glad to hear that you're enjoying the research feature — it's great to know it's making a meaningful difference for you.I truly appreciate your feedback on the text-to-speech functionality. Improving the voice quality and performance is definitely on my radar, and your input will help shape the next versions of the app.If you don’t mind, could you please submit a support ticket at oneai.framer.website/support? I would love to chat with you about improvements to the app.
Beautiful UI that under delivers
Huntsyea
It’s a beautiful app which attracted me to it, however I’m not paying for a beautiful UI when other free apps deliver 2x more functionality (e.g. the real value) and the same local models for free. They haven’t found their unique value so they are just throwing a paywall in a cash grab without thinking through what problem they are uniquely solving. Make all the local OSS models free (also add some of the heavier models a 16 pro can run) and build really good tools/integrations then run a freemium model on all the extra features and tools. Give me web search, cloud image models, connectors into Google Drive, Notion, etc. I will pay for that.
Ok but
LetsManuelENG
I guess it’s OK but most language models completely back out. If you ask them something more complex for an example I have got the high resource model installed on my device and the moment I ask you to speak German, it just completely breaks and spams Abu Dhabi. I guess it is quite a cool app, but it definitely needs some more improvement.
Developer Response
Thank you for the feedback. The German language issue with the "Abu Dhabi" spam sounds like a model routing bug - I'll flag this for our dev team to investigate.For more complex tasks and better language handling, you might want to switching to a more advanced model like Gemma 3 or Qwen 3 4B thinking, which tend to handle multilingual queries more reliably.Thanks for the constructive feedback - it helps us prioritize improvements.
Overall Positive
NuperForReal
Great app for just using offline LLM models for on to go research. UI/UX is really user friendly to use, like easy navigation, simple chatting, and a lot more. I recommend this as a more of a ChatGPT alternative with a similar UI.
Add Gemma 4 E2B and E4B on supported devices.
Version 0.5.2
The developer, Rica Harz, indicated that the app’s privacy practices may include handling of data as described below. For more information, see the developer’s privacy policy .
Data Not Linked to You
The following data may be collected but it is not linked to your identity:
Usage Data
Diagnostics
Privacy practices may vary, for example, based on the features you use or your age. Learn More
Accessibility
The developer has not yet indicated which accessibility features this app supports. Learn More
