PocketPal AI
How much does PocketPal AI cost?
Prices are current as of 9 Sept 2026.
PocketPal AI has no in-app products on record in the United States.
10+ alternatives to PocketPal AI
| PocketPal AI |   ChatLLM: AI Chat, Bot & Agent |   Locally AI by LM Studio |   Bria - VoIP Softphone |   Enclave - Local AI Assistant |   KDE Connect |   Pal Chat - AI Chat Client |   Venice AI |   Google AI Edge Gallery |   MLC Chat |   Google AI Edge Eloquent | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Rating / 5 | 4.1 | 4.8 | 4.7 | 4.4 | 4.7 | 4.3 | 4.3 | 3.6 | 4.0 | 4.3 | 3.7 |
| Reviews | 624 | 12K | 4.3K | 2.7K | 2.2K | 1.2K | 1.1K | 950 | 706 | 227 | 35 |
| Released | 2024 | 2024 | 2025 | 2016 | 2024 | 2022 | 2023 | 2025 | 2026 | 2023 | 2026 |
| Monthly price | — | $9.99 | — | — | $9.99 | — | $6.99 | — | — | — | — |
The going rate for each plan in United States, in its own currency, shaded from the cheapest price on the row to the dearest. Plans fewer than four in ten of the priced apps sell are left out.
PocketPal AI reviews
4.1 out of 5, based on 624 reviews.
What people like
- Totally free with no ads
- Beautiful interface and easy setup
- Works perfectly offline on your phone
What they complain about
- Context limit breaks conversations abruptly
- Poor formatting for code and math
- App crashes frequently on launch
Great App!
It doesn’t support AGI or cost money, doesn’t use wifi and is good?! Its perfect!
lies
the add lied it’s not free and limited convo based around token use big fat piers just like everyone else thanks for misleading me
Uh, I might have a small problem.
I may have triggered an accidental feedback loop by saying something specific to Muse. It fries their thinking.
v1.17.2
This new update fails to load any model but 2 stars for the effort.
Its bad but good
IT GOOD BEC OF HOW ITNWORKS OFFLINE BUT IT GIVES BAD AWNSERS THATS ALL U NEED TO KNIW GIVES DOG WATER AWNSERS
Best App for local ai on IOS, one thing.
Best app out there for local AI. huggingface integration is flawless. The only thing I ask is for there to be a way to allow web search when supported. Other than that, it's perfect. Easy to use and truly free.
Terrible!
Gives so many wrong answers and nothing current or live
How to disable reasoning?
I dont want it, it heavily slows down the response time.
iPhone17ProMax 频频闪退
完全没办法加 pal，一加就闪退，好多次直接把手机弄关机了
Stop iCloud model backups
Hey, can you stop backing up the models to iCloud?
Dude, Asghar, thanks a whole freaking lot! ^^
What?! No ads!🤯 WhaT WHaT!! Thank you!!! Full ability to get set up!🤓 Text to voice with 3 tiers 📣, and all free 🤘🏼🙏🏼 …and for myself, the only LLM Frontend that works on iOS 15.1 Big big thank you🙏🏼 you’re a bit of a wonderful anomaly of kindness and wellbeing passing some on to others 🖖🏼 One star reviews? Your competitors cause I can’t see how that’s decently possible. Please allow message to be sent with SHIFT+ ENTER without tapping or using a mouse.
System voice TTS not available
Despite the voice selection menu specifically saying that you can “pick a system voice — those are always available.” They aren’t available. Please fix.
Honestly one of the best but unreadable text
Out of all the apps I’ve been experimenting with on iOS for gguf models this is one of the best but the text formatting is terrible. There’s no bold, no italics or fonts. Also no line breaks or double spacing just endless run on sentence unreadable output. I’d really prefer this over most other stuff but that part is very frustrating. Please fix
Issues- seeking help
I have issues during chat after models loaded. No response, shows “0 ms/Token, 0 tokens/Sec”, can someone help me please? Thx
Good
I don’t have much space to take you
Suggestion
I’d like to see haptic feedback as an option and landscape mode could use some help. Otherwise it’s perfect.
Would be 5 stars if everything was labeled for VoiceOver
Please add VoiceOver accessibility. Button elements need labeling. Blind users deserve to be able to use this.
Good but missing an important feature
It’s really great of an app, I use it daily but it needs to make it so that it automatically delete any older messages in the chat when the context gets too big. I’m tired of having to start a new chat everyone I’m close to 8k context, and would prefer if the system automatically removed older chat messages in the conversation memory-that way it’s ongoing.
快捷指令无法使用
你这个快捷指令是不是有问题呀？
Very basic app that needs rethinking.
1. Instead of abruptly halting output mid-generation from running out of context tokens, deploy modern context-shift techniques where the app as a container summarises the previous key points into memory before truncation until next output. This alone reduces context size from 16384+ to 2048-1024 in the same session without breaking flow like with Qwen3.5 which overthinks. ↳ There seems to be no Context Size limit dedicated to ‘Thinking’ which can become an issue with models like Qwen3.5 that wants 128K tokens just for it even on 0.8 params. 2. Consider allowing multiple responses for continuation in the event one isn’t enough to contain the output of entire generation after hitting the token limit as one of the other efficient techniques especially for phones. 3. There’s no proper support for markdown (checkout ‘Markdown Quick Reference Cheat Sheet’ from WordPress) to paragraph spacing or diacritics if not text-symbols. 4. UX needs improvement like the fixed fields of System Prompt or the slider for Advanced Settings with a undo button which is very susceptible to unwanted touch from scrolling. 5. Not sure if it’s intentional but custom cores seem to reset after an app restart. Bonus: PocketAI is a better sounding name given the purpose unlike its very confusing affinity to Pixel Pals while having a blob emoji as its app icon, no wonder this is the last app people self-discover with flagships outside of AppStore when this is a very capable app.
Best app l have to say this one beats them all.
Beautiful UI, Great model support. I want to make heavy use in my shortcuts. The dev should simplify shortcut support. It’s very useful but shouldn’t require set up. I feel it should be streamlined and straight forward to get responses from models in shortcuts if you’re gonna implement it.
Not worth it
Probably the worst AI app i have ever used
Markdown
Assistant replies now not render formatting and equations. Markdown (headings, lists, code blocks, block quotes) and LaTeX math.
Bummer
Stopped loading models on iPhone 15 pro with the latest iOS.
Deepseek?
Hi I was told this app allowed for offline deepseek usage?
macOS Version ，Plz
I love this app，could you release a macOS Version ??
cannot use
iPados26 cannot download any of the model
Lacks Ministral 3 support but great otherwise
Only commenting to suggest that the latest version of llama.cpp is supported so I can use Ministral 3.
无法使用，一加载模型就闪退
It is unusable, and it crashes as soon as the model is loaded
stability issues?
In addition to some usability enhancements, it seems like the recent update introduced some stability issues. First, I only have 1 model loaded. But i’m often prompted to select a model to use. When I switch to another app to grab text to paste into PocketPal, the chat session does not always stay open and sometimes it asks me to select the model to use, so it’s like starting all over. if only 1 model is loaded, why doesn’t the app default to that model? I’m using this app for the privacy of offline use. I need to be able to reliably give it an instruction about the text it will receive and then go get that text from another app. It would be really helpful if I could save a set of instructions (such as to summarize structured meeting notes in a pre-defined format) to make it easier to use.
Great app, but needs a Context Sliding Window feature
I love PocketPal, but the app breaks when the context limit is reached. The AI simply stops responding to new messages instead of using a sliding window to clear old memory. Please fix this chat interruption issue and add a rolling context feature. It makes long conversations impossible.
Hey meta
Can we like be cool and you tell me how to create my own ❤️ or do you just want me to use your system like everyone else does😐
Context is full
"Context is full" will appear when processing larger text.
Unstable
Crashes often
Discrepancy Between Displayed and Actual Output Rates of Local AI Large Models; Seek Optimization
During the operation of large models on this local AI application, I've identified an issue that affects the user experience. The interface shows a character generation rate of over 50 characters per second. However, whether I employ large or small models, the actual output speed is noticeably slower, creating a significant gap from the indicated rate. Although local processing might be limited by device capabilities, this inconsistency still reduces efficiency. I sincerely hope the development team can look into this and make improvements to ensure the displayed rate matches the real - time output, thus enhancing the overall usability. Thank you for your efforts.
Seems to be one of the best out there
Hopefully it works with MLX soon. Having tried an increasing number of apps offering offline features, this one is one of the rarer ones that actually has uniquely good UI/UX that hopefully doesn’t change. And it has more neat things like the experimental pals feature. Downloading and perusing models is more polished than other apps too. It would be cool to have whisper models and other transcription models. Would be incredible to have TTS to somehow talk to the AI as well, with a little “hmm…” bubble as the response is processing.
crash after update
Keep crashing after update on iPadOS 26.4.
Very Good
Very good. Most of the Hugging Face models can be used as they are. It seems to offer the most freedom in terms of support. However, it would be great if it could support external APIs or iOS Shortcuts, so that questions could be sent directly to this app and responses received seamlessly.
为什么
为什么下载时进度条一点都不动
Improving VoiceOver Compatibility for a Better User Experience
I’ve been using this app for a while, and I truly appreciate its capabilities, especially the AI models available for summarization, rewriting, and instruction following. The responses are insightful and helpful, making it a valuable tool for various tasks. However, as a VoiceOver user, I’ve noticed that the app could be more optimized for screen reader accessibility. While basic navigation works, some elements are not fully labeled, and interacting with certain features can be challenging. Improving VoiceOver support would greatly enhance the experience for visually impaired users, making the app more inclusive and efficient. I highly encourage the developers to focus on accessibility improvements in future updates. Adding better screen reader support, clearer UI labeling, and smoother navigation would make this already great app even better for a wider audience. Thank you for considering this feedback!
Qwen3 thinking needs more tokens
Hey Dev, fantastic app. Can you allow more than 4096 output tokens? Qwen3 needs 16k to 32k just to think. Can you add an option for unlimited like llamacpp - set it to -1 I think?
Crashes
Crashes when I try to open app! Bravo on update guy!! iPhone 14 Pro Max iOS 17.2
Это хорошо, но есть нюансы
Я рекомендую это приложение, но есть несколько недочётов, на которые хочу обратить внимание: 1-лагающий интерфейс При использовании большой модели (у меня это LLAMA-META 8B) интерфейс приложения начинает фризить, у меня в один момент начали появляться артефакты, не дающие использовать софт 2-не загружается модель После установки модели, при попытке ее запустить софт может вылететь или просто не загрузить ее, как будто и не пыталась… как-то так, надеюсь, что поправят.
Loading issue
Model not getting loaded post any model installation. Device : iPhone 13/ios 18.2
Cannot load any model
Apple 14pro, iOS16.4.1 system, DeepSeek 1.5B Q4 cannot load the model, Qwen 0.5B cannot load the model, click to load the model, it flashes but nothing happens.
supports image input
this is amazing and free! thanks dev hope it supports image input too
Pobre experiencia
Lo he probado en un iPhone 8 y en un 16 Pro Max. Se instala bien pero se primero se congela intentando descargar un modelo de Hugging face (Qwen2.5 de 1.5 b) Luego se auto cierra y cada vez que haces clic sec abre y cierra al instante
Can‘t load model
why i downloaded the model，just press the load button，it didn't work 😂😂
A big problem
Always shows “ready for chat？load the last used model ”. Every model I downloaded from this app can not be used. And I delete the app then reinstall it for many times. iPhone 15 pro.
Ready to chat?
Hi I’m having trouble actually chatting with a model. I’ve downloaded a few with no issues, but when I load them it just says “ready to chat? Load the last used model”. I’ve done this over and over to no avail. Am I missing something? I’m on an IPhone X
PocketPal AI FAQ
- Is PocketPal AI free?
- Yes. The app is free, in all 175 countries it sells in, and it sells nothing inside it.
- Does PocketPal AI have a free trial?
- No. None of its offers (US) come with a free trial.
- What are the best alternatives to PocketPal AI?
- ChatLLM: AI Chat, Bot & Agent (4.8 from 12,177 reviews, premium from $9.99), Locally AI by LM Studio (4.7 from 4,347 reviews) and Bria - VoIP Softphone (4.4 from 2,735 reviews).
PocketPal AI download
Last updated 1 month ago.
PocketPal AI sells in 175 countries. It was released 2 years and 4 months ago, on 21 May 2024. It holds a 4.1 rating from 624 reviews.
Supports 1 language: English.
How much does PocketPal AI make?
Third-party estimates for the last 30 days, worldwide — not the store's own figures, which Apple publishes for nobody.