Layla v6.8.0 has been published
- Layla

- 7 days ago
- 2 min read
This update brings support for Gemma 4 - the latest model family from Google, improved agent functionalities, and German language support.
Gemma-4 Support
Gemma 4 is Google's latest open source LLM. It represents some of the best in class intelligence per parameter in local LLMs: https://deepmind.google/models/gemma/gemma-4/

Layla supports loading Gemma-4 GGUFs and LiteRT-LMs natively!
Layla will automatically use llama.cpp or LiteRT-LM inference engines depending on the models you choose.
Improved Agents - Choose Your Own Adventure
Layla Agents have improved with the ability to send progress messages back to the chat during execution. This allows your agents to generate UI options that can shape your chat after every message.
A demo is the "Choose Your Own Adventure" agent, where it will append options after every message.
Chat UI Customisation
You can now customise your chat screen to use traditional chat bubble style messaging:

You can change the display by going to Mini-apps -> UI Settings, scrolling down, and turn on "Use Chat Bubble Styles"
New Character Selection UI Layout
With more and more characters present in Layla, it's time we organised them into folders for easier access. This follow on the last folder feature update. Now the default characters are grouped together for ease of use. You can always create your own folders if you wish!

Full Changelog
New features:
added new Choose Your Own Adventure characters
supports the latest Gemma-4 model family (GGUFs and LiteRT-LM)
added German translations
Improvements:
added ability to send progress messages during python execution, you can view this progress in the Agent execution card to give you more insight into what the agent is doing
Layla Python can now generate files
new Python agents will automatically install needed dependencies when added for the first time
you can configure long-term memory to use a saved inference engine as summariser
added UI setting to display chats in a bubble-like interface instead (similar to Facebook Messenger or WhatApp etc.)
preset characters are now organised into folders
Bug fixes:
fixed a bug where sometimes character expressions will not be saved
fixed a bug where some memories cannot be processed
fixed a bug where CPU SD models cannot be run with LLM on the NPU
fixed a bug where attached inference settings are not applied when using a custom character as your phone's default assistant



Comments