Changelog - Ensu

A short summary list of changes to the Ensu mobile and desktop apps. For a more descriptive list with screenshots and blog post links, see the news.

v0.1.16 - May 2026

Added Gemma 4 for the desktop app.
Added local voice transcription for chat input on iOS and Android.
Improved image attachments with compression, thumbnails, full-screen previews, and cleaner composer previews.
Improved image prompt performance on native apps by resizing to model limits and caching multimodal context.
Added in-app changelogs to desktop, Android, and iOS.
Balanced haptics on mobile to once at the start of response generation.
Improved model downloads with shared native downloading, cache reuse, parallel/range downloads, retries, and better progress reporting.
Improved desktop model selection by using better system memory detection.
Fixed Android back button handling when the sidebar is open.
Fixed the desktop app version shown in settings.

Advanced model settings for power users: choose from preset models or bring your own.
Customizable system prompt on all platforms.
Smart default model selection based on your device's available memory (desktop).
Native inline math rendering on chat messages.
Model downloads now resume automatically after app restarts or interruptions.
Smarter retry behavior for interrupted assistant responses.
Smoother mobile chat experience with better keyboard and scroll handling.

Initial release of Ensu, Ente's local LLM app: a private, on-device AI that works offline. Cross-platform, open source, and powered by a shared Rust core.