SmolChat
io.shubham0204.smollmandroid
Zapstore _@zapstore.dev Republished from GitHub / F-Droid by the Zapstore main account.
SmolChat allows you to download and run popular LLMs on your Android device, locally, without needing an internet connection. Customize the model used for each chat, tune settings like temperature and min-p, and pin your favourite chats on the home-screen with shortcuts.
First release: Sep 19, 2025, 7 total releases.
Most recent release: Jun 21, 2026.
Appears in 1 app stack.
0 sats / 0 zaps received in the past year.
Sats Received
Underlying data available via MCP: app_zaps, app_releases.
Zap Count
Underlying data available via MCP: app_zaps, app_releases.
Releases
- Jun 21, 2026 16- Users can now enter multi-line queries without 'accidentally' sending the query to the LLM - Fix chat template parsing - Save system prompts and re-use them across chat sessions - Enable auto-scroll for chat messages
- Apr 17, 2026 15- Sync with upstream llama.cpp brings in support for Gemma 4 models (#128 and #130) - 'copy' and 'share' buttons below a message, now, do not copy the thinking contents of the model ( #126 by @Nittur) - Fix a bug where the app crashes when trying to show a list of available models from HuggingFace in the absence of an internet connection
- Mar 1, 2026 14- The app now supports automatic recognition (speech-to-text) for providing queries to the model. The speech-to-text capabilities are powered by [Moonshine](https://github.com/moonshine-ai/moonshine) and the packaged models could be found on [HuggingFace](https://huggingface.co/shubhxm0204/moonshine-asr-models/tree/main). - The speech is transformed into text during an on-device ASR model provided by Moonshine. Users will only require an internet connection when downloading the ASR models from the abovementioned HuggingFace Models repository. - Currently supported languages are English (`en`) and Chinese (`zh`). Although Moonshine supports more languages at the moment, SmolChat will add support for them in the future.
More…
- UI enhancements when downloading models from HuggingFace. Thanks to Gemini in Android Studio, the UI of the HuggingFace model browser screens has been improved and modernized. - A bug causing the app to crash after modifying model settings has been resolved (#114). Thanks to @jkkj for the contribution. - Feb 8, 2026 13- The app allows benchmarking the model used in a chat. Go to `Settings > Benchmark Model` in any chat and click 'Start Benchmarking'. The benchmark results are `pp` (tokens/sec for prompt processing) and `tg` (tokens/sec for token generation).
- - The app allows benchmarking the model used in a chat. Go to `Settings > Benchmark Model` in any chat and click 'Start Benchmarking'. The benchmark results are `pp` (tokens/sec for prompt processing) and `tg` (tokens/sec for token generation).
- - Add support for Portuguese language (alongside existing support for English and simplified Chinese). - Clear model context/memory when clearing the chat i.e. deleting all messages from the current chat. - Avoid model reloading when screen is rotated. ### UI Improvements - Earlier, the user had to click on the folder name to expand it. Clicking on the chevron icon now also expands the folder.
More…
- Task names are now appended with `[Task]` in the chats list to distinguish them from other chats. - - Fixed a bug where the app's memory usage kept increasing after switching models i.e. the memory acquired by the previous model was not 'released' when selecting a different model - Sync with upstream llama.cpp - Align default inference parameters with those found in `llama` executable ### UI Improvements - Chat message actions like share/copy/edit are now available in a dialog which appears when the message is long-pressed - Fix misleading/overflowing icons to enhance UX - Preserve query text in the search box when a model is opened while browsing HuggingFace