New native iPad U.I and Layouts. Introduce MLX on Device Inference. Improved Fllama Inference. Better Model Management for OnDevice Models. Added Dark mode icon Added Integration for iOS Shortcuts App. Added params for onDevice models Added more OnDevice Models to catalogue. Fixed iPad RAM detection issue Fixed Apple Pencil Integration. Fixed Community theme displaying twice. Default Context Size is now 16384 tokens. Search bars on model selection screens. Improved iOS and iPad OS Widgets. Introduced Widgets Settings Screen. Many more bugs and U.I Issue fixed.

Fixes missing providers options. Added on-device inference with MLX and Fllama. Added the ability to download HF on-device models. Ability to add attachments with support requests. Regenerating a response now saves all previous generations. Ability to bind memories to specific personas. Fixed Broken Pro Search. Fixed get_current_time tool call.

Added on-device inference with MLX and Fllama. Added the ability to download HF on-device models. Ability to add attachments with support requests. Regenerating a response now saves all previous generations. Ability to bind memories to specific personas. Fixed Broken Pro Search. Fixed get_current_time tool call.

Better Active and Loaded model state management. Better guide to Activate single Pro license on Multiple devices. Removed manual copying limitation of one paragraph. Fixed manual copying loosing markdown. Fixed Pro search MCP activation issue for some users.

Fixed Truncated fields on Model Parameters Fixed wrong error message about Max Output token Fixed Font Size not applying to some elements Fixed Copying message CRLF Mangling

Added ability to view attached photo from message bubble. Added Autonomous A.I Group Chat Added the ability to store chats on iCloud Fixed Copy button copying markdown with the original text Fixed A.I Generated Images disappearing after updates Fixed Themes not changing colors correctly on some places Fixed Group chats not talking in second person view to user Added more group chat settings in chat window

Fixed unable to copy more than one paragraph from response bubble Removed Pro Search fallback from SearXNG

Fixed SearXNG not working for some user. Fixed MCPs compression when there are too many active MCPs.

Fixed U.I Glitching on Image Attachments Fixed App Crashing when processing Images Fixed wrong modal loading context parameter Removed tool calling from memory extraction

Added Custom Headers Support for Server Connectivity Negative prompt Field on ComfyUI is now Optional Fixed Missing Windows Link for LM Connect Updated Docs Updated Translations

Added More Model Parameters to Global Settings Added toggle to change Keyboard Return to Send or New Line Added Comfy UI integration (beta) Added USB Connection mode. Fixed LM Connect Download button Fixed Open router vision model detection.

Added Personas Added Group Chats with multiple models Added Import chart from LM Studio to LM Mini Fixes SearXNG button not working with Pro Search Fixes Remote URL for A1111 not working

Fixes Memory Update tool. Pro Search is prioritized over other search tools. Fixes theme reapplying on restarting the app. Fixes conversation started from folder won't appear in folder. Feature Request now mark Unread replies. Fixes chat backgroun sometimes disappearing on their own. Auto retry if Model ends response abruptly.

Use Kokoro on PC with LM Mini Connect Introduced Theming to really make app vibe with you Community theme: Share or Download themes generated by Other users. Conversations now Display Model used and Last used time. Added Section to Create Custom Themes Fixes bug where Voice mode acted as Call Mode

Google Login fixed Added Chat icon size controls under appearance New Call Mode (Beta) Encrypted remote access Added chat management features better model loading progress Change logs section added Font family feature added to chat customization Improved analytics dashboard Welcome wizard for new users Improved Support / Feature Request Commenting system

Fixes Pro Search

Added Remote Access via LM Connect on Mac Added Live Activity Widget Added Font Size settings Fixed Stats not showing Fixed broken SearXNG Fixed broken Tool Call

multilingual text fixes.

- Fixed Cloud Providers for Pro Users - Improved Memory extraction tool for Pro Users - Fixes bug where thinking mode wouldn't show thought process - Fixes bug where per conversation model selection was broken - Added Ability to select providers for each conversation - Ability to load images from Files App. - Backups not store memories too, using same encryption. - Improved Premium Search for faster results.

Image Generation and Image Prompts using Automatic1111. Ability to Auto generate Images for every response with AI Generated Image Prompts. Ability to attach images from Files App. Fixes bug where chat losses Context. Fixed Math Formulas from different model Added more tags to Thinking extraction. Pro Tier Introduced for Native web search without MCP, Cloud Backups, Conversation Branching and Encryption.

Fixes chat context for older chats. Fixes crash on incomplete responses. Better handling for Mathematical Formulas. Fixes MCP missing chat context. Fixes Local Network permissions not found. Returns correct error from LM Studio on failed generation. Fixes tools badge overflow on response bubble. Fixes cluttered view for tool call result json view.

Better UI for Chat message field. Fixed load model name overlapping copy icon on chat bubble STT now understands languages other than English Screen timeout is blocked when long audio message is playing. Better Tool Calling Loop to avoid hitting tool calling limit.

Added on Device Voice Mode Using TTS and Kokoro On device model.

App is now Multilingual Response can now be regenerated Responses are now editable MCPs are prioritized over SearXNG and Search API Some U.I bugs fixed.

Ability to store Multiple System prompts and assign them per chat Ability to bind system prompts with Models Fixed long message breaking the Input field Fixed not able to create new line in input field