Added support for the new Gemma 4.
Available in the model manager:
- Gemma4 E2B + vision
- Qwen3.5 4B + vision

Disabled automatic reasoning for some models that reason by default.

Fixed an issue with the remote chat that prevented the connection with certain providers.

Qwen 3.5 is now available in the model catalog (chat and vision).
Updated compatibility with the latest models.
Import mmproj files for your custom models that support vision capabilities.

Added support for Anthropic-compatible API endpoints in the remote chats.

Processing input images for local chats is now MUCH faster.
Fixed some inputs parameters for the remote chat.
General UI improvements for the chat interface.
Improved visualizations of tables, code blocks, lists and headers.

Improvements for remote chats!
- You can now use the LLM Client to send images using multimodal vision models
- Set additional chat parameters for OpenAI compatible APIs and Ollama
- New voice mode: talk to start a chat and listen to the response
- Introduced a selection of themes for the client chat interface

General improvements:
- Fixed an issue that prevented text from being added to RAG from transcripts
- Minor improvement in speech synthesis

Transcription has been completely revamped!
- Multilingual transcription supported with different Whisper models.
- Transcribe audio files or live recordings.
Improved RAG management and navigation, added support for html and json files.
You can now customize settings for Shortcuts separately from the main app.
More UI improvements across the whole app.

Feedbacks or questions? info@latentlake.com

Eyes for your LLMs: 
- Now you can try vision models to process images!
- Added Qwen3VL, InternVL and LFM2.5 VL to the model catalog.

Improvements made after your feedback:
- Switch models and adjust parameters easily without leaving the chat.
- The model catalog has been revamped with additional info.

For more suggestions, info@latentlake.com.

- Added LFM2.5 Thinking (by Liquid) in the model manager
- Improved UI for reasoning models
- New feature: Search across chats and messages
- Cancel an ongoing request for remote chats (requires server support)

- NEW MODELS: Added Ministral 3 and LFM 2.5 (by LiquidAI) to the model manager.
- BACKUP & RESTORE: Manually backup, export and import your chats.
- Updated support for the latest models.
- Improved the chat UI with a better font and more space for messages.
- Transcribed text is now added to the text input field.
- Reduced app size by making the Whisper model downloadable.

- Remote chats get additional features such as tags and pinning.
- Long-press the (+) button to switch the chat type from local to remote.
- Added option to save speech audio from chat messages
- Text to speech settings now persist across the app (bug fix).

Feedbacks or questions? info@latentlake.com

A major update is here with a new feature: Text to speech!
-  Read your chat messages aloud
- Try different voice styles 
- Supports English, Spanish and French
- Export audio speech from long texts

Important fixes:
- Fixed an issue that could cause the app to crash when using shortcuts
- Dark/Light mode now automatically switches when system settings change
- General UI improvements

feedbacks or questions? info@latentlake.com

Fixed an important bug that caused the app to crash with local models.
- improved RAG topic selection for local models
- general UI improvements
- now you can re-generate the last message in the LLM Client

feedback or questions? let me know at info@latentlake.com

Some improvements for the LLM Client:
- now you can re-generate the last message
- some UI improvements

feedback or questions? let me know at info@latentlake.com

LatentChat is now much lighter!
You can delete the default model from the model manager.

Other minor improvements.

Feedback or questions? info@latentlake.com

Fixed an issue for the port field in the remote chat connection.

Now you can select and delete multiple chats at once.
Resolved a bug with answer generation on retry.

Added Qwen3 4B in the model manager.

Added support for Gemma 3n and other recent models.
Better memory management when swapping models.

Some interesting improvements with this update!

- Continue a remote chat even after closing it
- Copy a remote chat to be used with on-device models
- Text files can now be added in the remote chat
- Improved markdown for code, html or tables
- Remote chats can now be branched

Fixes:
- solved an issue that prevented to send a message with a long input
- added a confirmation before deleting a chat
- fixed ollama connection

Fixed an issue with some models that failed to load.

Qwen3 added to the model manager!

Now you can view HTML generated pages directly from the chat.
Improved remote llms:
- the latest selected model will persist between sessions
- better error handling

Added support for OpenRouter in the remote chat.
Minor bug fixes.

Now you can use Gemma 3 1B and 4B!
Solved an issue with the advanced parameters not being persistent.
Added Qwen2.5 3B in the download manager.