Home › Utilities
US
1 of 1 storefronts
Content in en-US
Content in en-US
Information
Preview 2 platforms
iPhone 10 images
iPad 8 images
What’s New Version 4.4.10
[New] Added parallel inference support for Apple MLX
[New] You can now export all conversation data from Settings to SQLite or Markdown + JSON
[New] Added NPU acceleration support for the Qualcomm Snapdragon X2 Elite chip
[New] RWKV can now answer different questions at the same time in the response style selection panel
[New] Added support for more Qualcomm Snapdragon series chips
[Improved] Improved how image messages are displayed in visual understanding
[Improved] Improved CoreML loading speed on Apple chips
[Improved] Optimized display text for some thinking modes
[Improved] Batch Inference now supports Pause/Resume
[Fix] Fixed an issue where the download button could disappear in some cases
[Fix] Fixed an issue where the download button could disappear in some cases
[Fix] Fixed an issue where weights could be automatically loaded twice in some cases
[Fix] Fixed an issue where conversation logs could not be exported in some cases
[Fix] Fixed exception log counting during conversation export
Description
App Name
RWKV Chat - Local AI
Subtitle
Fast, private, and local AI chat assistant.
Promotional Text
Experience the power of RWKV models directly on your device. Completely offline, privacy-first, and efficient. No internet required.
Description
Unlock the power of Large Language Models right in your pocket.
This app is a lightweight, privacy-focused AI chat client powered by the revolutionary RWKV architecture. Unlike traditional cloud-based AIs, this app runs entirely on your device, giving you a powerful intelligent assistant that works anywhere, anytime—even without an internet connection.
Why choose this app?
1. Total Privacy
Your data never leaves your device. All processing happens locally. There are no servers, no cloud uploads, and no tracking. You have complete control over your chat history.
2. Powered by RWKV
Experience the unique advantages of the RWKV model. It combines the high performance of Transformers with the inference efficiency of RNNs. This results in faster generation speeds and lower memory usage compared to similar-sized models.
3. Offline Capability
Perfect for travel, remote work, or simply when you want to disconnect. Your AI assistant is always ready to help you draft emails, brainstorm ideas, translate text, or just chat, regardless of your internet connection.
4. Optimized Performance
Designed specifically for mobile hardware to ensure smooth, low-latency conversations without excessive battery drain.
Key Features:
- Intuitive Chat Interface: Clean design for seamless conversations.
- Model Management: Easily switch between different model sizes to balance speed and intelligence based on your device capabilities.
- Customizable System Prompts: Configure the AI to act as a coder, creative writer, translator, or general assistant.
- History Management: Save, view, and delete your conversation history locally.
About RWKV Technology:
RWKV (Receptance Weighted Key Value) is a novel RNN architecture that delivers Transformer-level LLM performance. It allows for linear attention during inference, making it significantly more efficient for local deployment on consumer devices compared to traditional architectures.
Download today and carry a personal AI assistant in your pocket.
Accessibility
ipadiphonemacrealityDevicetvoswatch
Version history 14 versions
Build 883460273v4.4.102026-05-16 17:43:35
[New] Added parallel inference support for Apple MLX
[New] You can now export all conversation data from Settings to SQLite or Markdown + JSON
[New] Added NPU acceleration support for the Qualcomm Snapdragon X2 Elite chip
[New] RWKV can now answer different questions at the same time in the response style selection panel
[New] Added support for more Qualcomm Snapdragon series chips
[Improved] Improved how image messages are displayed in visual understanding
[Improved] Improved CoreML loading speed on Apple chips
[Improved] Optimized display text for some thinking modes
[Improved] Batch Inference now supports Pause/Resume
[Fix] Fixed an issue where the download button could disappear in some cases
[Fix] Fixed an issue where the download button could disappear in some cases
[Fix] Fixed an issue where weights could be automatically loaded twice in some cases
[Fix] Fixed an issue where conversation logs could not be exported in some cases
[Fix] Fixed exception log counting during conversation export
Build 883391308v4.2.62026-03-15 22:18:08
Added “RWKV Ask for You” to generate questions from chat context and prefixes.
Added MLX support for RWKV7-G1e 7.2B on macOS.
Rebuilt the bot interaction panel with runtime stats (prefill, decode, tokens).
Added conversation token limit reminders.
Refactored remote config and update logic.
Improved response speed and generation quality.
Added model loading progress display.
Improved RWKV voice generation experience.
Optimized chat input, message layout, and navigation.
Various stability fixes and minor UI improvements.
Build 882468996v4.0.72026-02-14 20:20:51
[New] Brand-new UI upgrade: a cleaner overall look, more consistent buttons and layout, and more intuitive operations
[New] Added an Apple-specific inference acceleration solution for iOS devices, greatly improving performance
[Fix] Fixed a freeze issue when opening the weights storage folder on some Windows devices
[Fix] Fixed an issue where the model could unexpectedly stop outputting when generating User
[Fix] Fixed abnormal application language after displaying open-source licenses
Build 882178056v3.9.62026-02-10 20:45:10
[New] RWKV Chat can now use any .pth file on your computer
[New] You can now freely manage the storage location of RWKV Chat weight files
[New] You can now set the application font (currently ColorOS has font display issues, which we are fixing)
[Improved] Optimized RWKV-VL inference performance
[Improved] We've updated to a more fun app icon
[Improved] Improved system prompt settings interaction, #62, by @yuyi2439
[Fix] Fixed several minor issues
Build 881930458v3.8.12026-01-28 19:24:06
[New] Added the brand-new RWKV-VL model
[New] We added more detailed and accurate release notes
[New] You can now set the application font (currently ColorOS has font display issues, which we are fixing)
[Improved] Performed several performance optimizations
Build 881538324v3.6.62026-01-15 23:45:49
* [New] Added model file import and export functionality for RWKV Chat.
* [Optimization] Downloaded weights are now stored in the application's sandbox directory under the /rwkv_chat_models folder instead of the system Documents directory (Windows, macOS).
* [Optimization] Fine-tuned decoding parameters for Conservative Mode.
* [Optimization] Improved the UI and copywriting for the Model Management page.
* [Fix] Corrected the button position in Completion Mode.
* [Fix] Fixed several minor bugs and improved stability.
Build 881369920v3.5.62026-01-10 23:15:58
- [New] Added File Management to the Settings page, allowing for centralized management of downloaded weight files.
- [Optimization] Integrated Performance Benchmark and Lambada Test functions into the Homepage.
- [Optimization] Lambada Test now supports rendering model output results.
- [Optimization] Refined the descriptions for Android NPU support.
- [Optimization] Improved Inference Engine performance.
- [Fix] Fixed multiple known bugs.
Build 881303402v3.5.02026-01-08 13:52:00
[New] Added File Management to the Settings page, allowing for centralized management of downloaded weight files.
- [Optimization] Integrated Performance Benchmark and Lambada Test functions into the Homepage.
- [Optimization] Lambada Test now supports rendering model output results.
- [Optimization] Refined the descriptions for Android NPU support.
- [Optimization] Improved Inference Engine performance.
- [Fix] Fixed multiple known bugs.
Build 881050726v3.3.42025-12-29 00:58:10
- We have fix some minor issues in this version.
Build 880966330v3.3.12025-12-24 09:59:23
We have fixed some minor bugs in this version.
Build 880551582v3.2.12025-12-11 19:25:59
Added: Windows now supports using the Albatross plugin for inference.
Added: Offline translation now supports OCR (photo) recognition.
Added: System prompt templates now support inserting dates.
Improved: Optimized code-block rendering.
Improved: Optimized certain UI display elements.
Improved: New “continue writing” mode UI and interactions.
Build 880247282v3.1.102025-12-01 21:40:19
1. Revamped the Completion page UI to make it cleaner and easier to use.
2. Fixed a critical bug causing crashes on iOS 17 devices.
3. Made minor UI/UX improvements.
Build 880223046v3.1.82025-12-01 17:12:15
Improved: We've revamped the user interface (UI) of the Completion page to make it cleaner and easier to use.
Fixed: Squashed a critical bug causing the app to crash on iOS 17 devices.
Build 871551573v1.02025-11-26 20:03:03
No pricing data captured yet — comparisons appear once the app has paid storefronts or in-app purchases.
Availability 1 of 1 storefronts
| Region | Language | Price | Ratings | Avg | Version |
|---|---|---|---|---|---|
| US | en-US | Free | 0 | — | 4.4.10 |
Change log 17 changes · US
✎
releaseNotes updated
[New] Added parallel inference support for Apple MLX
[New] You can now export all conversation data from Settings to SQLite or Markdown + JSON
[New] Added NPU a
✎
description updated
App Name
RWKV Chat - Local AI
Subtitle
Fast, private, and local AI chat assistant.
Promotional Text
Experience the power of RWKV models directly on your devic
✎
name updated
RWKV Chat
◆
Version 4.4.10 released
[New] Added parallel inference support for Apple MLX
[New] You can now export all conversation data from Settings to SQLite or Markdown + JSON
[New] Added NPU acceleration support for the Qualcomm Snapdragon X2 Elite chip
[New] RWKV can now
◆
Version 4.2.6 released
Added “RWKV Ask for You” to generate questions from chat context and prefixes.
Added MLX support for RWKV7-G1e 7.2B on macOS.
Rebuilt the bot interaction panel with runtime stats (prefill, decode, tokens).
Added conversation token limit rem
◆
Version 4.0.7 released
[New] Brand-new UI upgrade: a cleaner overall look, more consistent buttons and layout, and more intuitive operations
[New] Added an Apple-specific inference acceleration solution for iOS devices, greatly improving performance
[Fix] Fixed a
◆
Version 3.9.6 released
[New] RWKV Chat can now use any .pth file on your computer
[New] You can now freely manage the storage location of RWKV Chat weight files
[New] You can now set the application font (currently ColorOS has font display issues, which we are fi
◆
Version 3.8.1 released
[New] Added the brand-new RWKV-VL model
[New] We added more detailed and accurate release notes
[New] You can now set the application font (currently ColorOS has font display issues, which we are fixing)
[Improved] Performed several perform
◆
Version 3.6.6 released
* [New] Added model file import and export functionality for RWKV Chat.
* [Optimization] Downloaded weights are now stored in the application's sandbox directory under the /rwkv_chat_models folder instead of the system Documents directory (
◆
Version 3.5.6 released
- [New] Added File Management to the Settings page, allowing for centralized management of downloaded weight files.
- [Optimization] Integrated Performance Benchmark and Lambada Test functions into the Homepage.
- [Optimization] Lambada Tes
◆
Version 3.5.0 released
[New] Added File Management to the Settings page, allowing for centralized management of downloaded weight files.
- [Optimization] Integrated Performance Benchmark and Lambada Test functions into the Homepage.
- [Optimization] Lambada Test
◆
Version 3.3.4 released
- We have fix some minor issues in this version.
◆
Version 3.3.1 released
We have fixed some minor bugs in this version.
◆
Version 3.2.1 released
Added: Windows now supports using the Albatross plugin for inference.
Added: Offline translation now supports OCR (photo) recognition.
Added: System prompt templates now support inserting dates.
Improved: Optimized code-block rendering.
Imp
◆
Version 3.1.10 released
1. Revamped the Completion page UI to make it cleaner and easier to use.
2. Fixed a critical bug causing crashes on iOS 17 devices.
3. Made minor UI/UX improvements.
◆
Version 3.1.8 released
Improved: We've revamped the user interface (UI) of the Completion page to make it cleaner and easier to use.
Fixed: Squashed a critical bug causing the app to crash on iOS 17 devices.
◆
Version 1.0 released