PocketPal AI
仕事効率化
無料 · アプリ内購入 · iPadに対応しています。macOSでは検証されていません。
Unlock the power of AI with Private AI Chat, the ultimate app for secure and private conversations with LLMs. The app brings the cutting-edge AI directly to your device, ensuring your chats remain confidential and offline.
Key Features:
- Offline Chatting: Interact with advanced AI models directly on your device without the need for an internet connection.
- Hugging Face Integration: Search for GGUFs on Hugging Face directly within the app, bookmark or download your favorites.
- Secure & Private: Your data never leaves your device. All conversations are processed locally, providing security and peace of mind.
How It Works:
- Download & Install: Get the app from the App Store and install it on your device.
- Download Model Weights: Connect to the internet to download the required model weights (in GGUF format, e.g., from Hugging Face).
- Chat Offline: Once setup is complete, enjoy private, offline conversations with the model anytime, anywhere.
さらに見る 他のアプリでは選択できるモデルがプリセットであったり条件がついていたりするが、このアプリはハードウェアのスペックが許す限り無制限である。ただしあまり大きなモデルをセットすると、動作が重くなるし現実的ではなくなってくる。それでもその自由性があるおかげで、このアプリはお話相手やプログラム相談に使うには不自由しない。RAG(ファイルを読み込ませて分析をする)は使えないし、アシスタントやロールプレイはまだ実験的の表記で完全には実装されていないので、このアプリひとつでLLMでやれることをなんでもこなせるというわけにはいかないが、AIって馬鹿だなあという不満を感じることがあるなら試す価値はあると思う。Vecteus-v1なんかを読み込ませれば、いい感じの話し相手になってくれます。
他のアプリでは選択できるモデルがプリセットであったり条件がついていたりするが、このアプリはハードウェアのスペックが許す限り無制限である。ただしあまり大きなモデルをセットすると、動作が重くなるし現実的ではなくなってくる。それでもその自由性があるおかげで、このアプリはお話相手やプログラム相談に使うには不自由しない。RAG(ファイルを読み込ませて分析をする)は使えないし、アシスタントやロールプレイはまだ実験的の表記で完全には実装されていないので、このアプリひとつでLLMでやれることをなんでもこなせるというわけにはいかないが、AIって馬鹿だなあという不満を感じることがあるなら試す価値はあると思う。Vecteus-v1なんかを読み込ませれば、いい感じの話し相手になってくれます。
Thanks Qwen3 for the quick support and Japanese translation!
Thanks Qwen3 for the quick support and Japanese translation!
One feature request: please add hardware keyboard support on iPad.I use iPad Pro with Apple Magic Keyboard, but pressing Enter only creates a new line instead of sending the message.It would be great to have:* Enter → Send* Shift + Enter → New line or an option to customize this behavior.This would make PocketPal much more comfortable for iPad users.
One feature request: please add hardware keyboard support on iPad.I use iPad Pro with Apple Magic Keyboard, but pressing Enter only creates a new line instead of sending the message.It would be great to have:* Enter → Send* Shift + Enter → New line or an option to customize this behavior.This would make PocketPal much more comfortable for iPad users.
こちらの環境はiPhoneAirですがHugging Faceから追加したQwen3.5-4B系列のローカルモデルが問題なく動作してます。アプリは日本語対応、モデル設定からカスタム指示も入れられて良きです。
こちらの環境はiPhoneAirですがHugging Faceから追加したQwen3.5-4B系列のローカルモデルが問題なく動作してます。アプリは日本語対応、モデル設定からカスタム指示も入れられて良きです。
- New: support for LFM3 and Bonsai 2 models.
- New app icon, with dark and tinted versions.
- Faster text generation on iPhone and iPad.
- Fixed: hands-free voice mode could fail without internet for some voices.
- Updated translations.
v1.22.0 2日前
- Hands-free voice mode: talk to PocketPal (with PocketPal Plus).
- Voice input: tap the mic and speak instead of typing.
- New voice, Phonon: a natural English voice (iOS 18+).
- PocketPal Plus: an optional one-time unlock to support development.
- Web search with You.com, no API key needed.
- Bug fixes and stability improvements.
v1.21.2 10月27日
New in PocketPal:
• Sign in with Apple
• Arabic app language
• Scan a QR code to add a server (remote llama.cpp)
• Load and unload llama-server models from your phone (remote llama.cpp)
• Firecrawl as a web search option
• Delete your account in Settings
Fixes:
• A crash when a model was unloaded mid-reply
• The keyboard covering fields in sheets
• Stop now cancels a running web search
• Cleaner chat bubbles and read-aloud when tools run
v1.20.1 10月20日
Buy Pals directly in the app
v1.19.1 10月8日
- Faster on Snapdragon's AI chip (NPU). Many more models now run on the NPU, including popular Q4_K_M and Q6_K downloads, with up to 10× faster prompt processing and noticeably quicker replies.
- Spanish is here. PocketPal is now available in Spanish.
- Updated AI engine with the latest llama.cpp improvements.
- Accurate context meter for remote servers.
v1.18.0 9月28日
- Added Spark2.5 model support
v1.17.3 9月11日
• llama.cpp capability discovery: PocketPal now detects a connected server's capabilities, e.g. vision support and context size, detects and warns when a conversation runs out of room.
• Internet search in chat: Pals can search the web and read pages. Bring your own API key (Brave, Tavily, Exa). Your key stays on your device.
• Speculative decoding (experimental): pair a small draft model for faster replies.
• Pin chats to keep them at the top of your list.
• Export a chat as Markdown.
• Polish and Portuguese (PT) added, plus a searchable language picker.
• Updated llama.cpp engine: newer model support and performance improvements.
• Fixes: imported models no longer disappear after an iOS update or restore.
v1.17.2 8月28日
- Improved onboarding experience .
- Smarter AI model selection.
- Add 31 languages for the text-to-speech model (Supertonic)
- Added Brazilian Portuguese support.
- Improved download reliability.
v1.16.1 7月14日
Warn when a chat is near or at the context limit.
v1.15.2 6月8日
• Added Render HTML, Date & Time, and Calculator tools
• Enhanced text-to-speech and fixed an issue that could cause silent Kokoro audio
• Includes several improvements and bug fixes
v1.15.1 5月30日
• Neural TTS with streaming: 3 neural engines (Supertonic, Kitten, Kokoro), voice-led setup
• New model support via llama.cpp b8827: Gemma 4, MERaLiON-2, Bonsai
• Ukrainian & Traditional Chinese UI
• Unlimited generation length (n_predict default lifted)
v1.14.0 4月21日
- Added Gemma 4 support
- Squashed a few bugs
v1.13.3 4月8日
- Remote model support: connect to OpenAI-compatible servers (LM Studio, Ollama, etc.)
- Table rendering in Markdown responses
- New languages: Farsi, Korean, Malay, Russian, Hebrew, Estonian
- Fix: thinking blocks no longer disappear after completion
- Fix: model dropdown scrollable on smaller screens
- Fix: thread count preserved across restarts
v1.13.1 3月28日
- Added Tiny Aya support
- Improved chat management: Bulk delete and export chats
- Better vision model control: Configurable image token limits
- Added Indonesian
- Fixes
v1.11.21 2月17日
- Added support for renaming models
- Improved performance and fixed a few bugs/crashes
v1.11.16 1月26日
- Added support for renaming models
- Improved performance and fixed a few bugs/crashes
v1.11.15 1月26日
A few bug fixes.
v1.11.13 1月14日
- Add Qwen3-VL and Ministral 3 support
- Bug fixes
v1.11.11 2025/12/22
- Adds Shortcuts
- Made some UI enhancements
v1.11.7 2025/11/04
- Added integration with Palshub.ai
- Performance enhancements and numerous other improvements under the hood
v1.11.2 2025/10/10
- Added LFM2-VL models support
v1.10.24 2025/08/21
- Import/Export Pals: Easily back up or share your pals!
- GPT-OSS Support: Enabled! May your device have enough RAM... and courage.
- Report Content: Now you can report inappropriate content.
- Bug Fixes.
v1.10.19 2025/08/13
- Added support for LFM2 models
v1.10.16 2025/07/29
- Added thinking control
v1.10.15 2025/07/13
- Added SmolLM3 support
v1.10.14 2025/07/10
- New: support for LFM3 and Bonsai 2 models.
- New app icon, with dark and tinted versions.
- Faster text generation on iPhone and iPad.
- Fixed: hands-free voice mode could fail without internet for some voices.
- Updated translations.
さらに見る バージョンv1.22.0 2日前
データの収集なし デベロッパはこのアプリからデータを収集しません。
