Spike AI

- Introduced the new Card tool for richer, structured results directly in chat. - Hardened Agent Mode for more reliable tool handling and a cleaner execution flow. - Improved tool and thinking timelines so intermediate steps are easier to read. - Fixed a chat UI issue where tiny empty thinking entries could appear after tool responses. - General stability improvements and bug fixes.

Spike Agent is here: let Spike plan, use tools, and work through tasks step by step on a connected server.

Redesigned side menu plus small bug fixes and improvements.

• Rebranded to Spike — new icon, ghost, and Orbitron sidebar • Smoother thinking box: faster open, crisper edges, no more double-toggle • Live thinking preview during tool chains • Native quick menu and refreshed sidebar spike icon • Improved scroll behavior and streaming gap fixes • Sharper Qwen 3.6 vision quality

- Fixed tool cards (crypto, maps, etc.) disappearing in chat - Added Unsloth models to Hub search - Bug fixes and stability improvements

- Apple Watch app — Chat with your AI directly from your wrist (server models) - On-device Web Agent — Browse and interact with websites autonomously - Welcome screen — Randomized color palettes with better color separation - Ghost Mode pill button — Ghost toggle and new chat merged into a compact combined button - Glass UI — Tool chips above the input bar now use clear glass instead of filled accent color - Header layout — Model picker no longer overlaps side buttons; larger tap targets throughout - Sidebar — Chat history scrolls to the very bottom of the screen - Scroll stability — Reduced jitter during tool call transitions - Server reliability — Improved connection handling and error recovery - Gist sharing — Published artifacts are now secret/unlisted - Bug fixes — Fixed streaming issues, Watch layout, TTS cleanup, and scroll-to-bottom edge cases

- New: Apple Watch app - Bug fixes and improvements

- Unified Thinking & Tool Timeline: Thinking and tool calls now appear in a single collapsible timeline card. Tap to expand and see every step of the AI's reasoning and actions live — from thinking to web searches, memory lookups, and more. - Streaming Image Generation: Watch images appear progressively as they're generated, with a live preview that builds up in real time. - Improved Stability: Fixed connection issues after returning from standby, prevented out-of-memory crashes when switching between models, and increased connection timeout for slower model loading. - Better Scrolling: Fixed scroll position jumping when dismissing code or HTML sheets, and improved auto-follow behavior during streaming. - Premium Markdown: Redesigned tables, cleaner code block toolbars, and smoother rendering throughout. - UI Polish: Balanced message padding, default accent color set to purple, and various visual refinements.

- Now supports Qwen 3.5 models - Redesigned markdown tables - Visual polish and bug fixes

- Smart Home Control — Control all your Apple Home devices with natural language. Turn lights on/off, set brightness, change colors, run scenes, and more. Control multiple devices at once by room or name. - Apple Maps — Search for places, get directions, plan multi-stop routes, and view locations on an interactive map — all through conversation. - Soprano TTS — New on-device text-to-speech engine replacing ElevenLabs and System TTS for faster, more natural voice output. - Improved Whisper — More reliable on-device speech recognition with better error recovery. - Model Selector — Long model names now truncate gracefully instead of breaking the toolbar layout. - Native Tool Calling — More reliable tool execution with better parameter handling across all tools.

What's New in 1.4.6 - New Reminders Tool: Create and manage reminders with batch support, full list management, and a beautiful Apple Reminders-style UI - New Calendar Tool: Add and view calendar events with quick toggles, syncing directly to your Apple Calendar - New Crypto Tool: Track cryptocurrency prices with a dedicated UI and live data - On-Device AI: Optimized local MLX inference for faster performance on iPhone - Camera improvements: Better image sizing when capturing photos - Improved markdown rendering for crypto and math content - Various bug fixes and performance improvements across the app

New: Artifact preview for generated HTML code blocks. View, copy, and navigate between artifact versions directly in the app.

Bug fixes and minor optimizations.

Bug fixes and minor optimizations.

Bug fixes and minor optimizations.

Bug fixes and minor optimizations.

Bug fixes and minor optimizations.

Bug fixes and minor optimizations.

• Local mode has been added to enable the use of MLX models on the device

• A new design that matches iOS 26

• Bug fixes and overall stability improvements • Added favorite menu

• Bug fixes and overall stability improvements • Added light mode

• Added a new image generation tool powered by a custom backend • Bug fixes and overall stability improvements

• Bug fixes and stability improvements

• New Voice Mode • Bug fixes and stability improvements