Changelog

What's New in Lexika

Stay up to date with the latest features, improvements, and bug fixes. We're constantly working to make Lexika better for you.

v4.1.0

Current version

A New Look, Three New Languages, and Support Inside the App

Lexika wears its new Signal Field identity, speaks Spanish, Japanese and Vietnamese, and gets a rebuilt pricing page, a model finder, in-app support tickets, and a long list of polish on every platform.

  • Lexika has a new look: the Signal Field identity brings a refreshed palette, animated avatars for people and workspaces, and a favicon that follows your theme
  • A rebuilt landing page with a story-driven demo of a real conversation, new sections for teams and organizations, and the latest from the blog
  • The pricing page is rebuilt as a plan ladder in the same design, and plan cards now say how many Brains and entries each plan includes
  • Start the free trial straight from a plan card when your account qualifies
  • Spanish, Japanese and Vietnamese join English and Arabic — the app, the Knowledge Center, this changelog, and the model and plan descriptions are all translated
  • Pick a violet surface palette in Settings, independently of light or dark mode
  • Choose a chat background in Settings — eight patterned grounds or a solid one
  • A model finder in the models gallery: answer three questions to get a recommendation, then use it in chat or open its spec — also reachable from the chat's model selector
  • Drag your favorite models to reorder them
  • Rebind any global keyboard shortcut from the Shortcuts tab in Settings
  • Download a code block straight from its header
  • Get help without leaving Lexika: open a support ticket from Settings, and every contact link now leads to the support page
  • A transactions page under Pricing lists your full purchase history
  • Usage in Settings can be viewed step by step or grouped by feature, with a total row
  • Choose a background for the Brain map — dots, a grid or a constellation
  • Smooth follow-scrolling while a reply streams can be switched off in Settings
  • Models that are temporarily down are marked unavailable, and a failed send explains itself in plain language
  • Streamed replies keep up with bursts, and the model selector remembers which rail you were on
  • Workspace invitations, folder moves and Brain links now update live across your devices
  • Lexika asks before deleting a conversation, a lab chat or an attachment, and Shift-click selects a range in the attachment picker
  • Laboratory chats start from starter cards, and the Agentic Research setup reads as one editable sentence
  • Members of a workspace no longer see Brain write controls they cannot use, and the sidebar hides Brains and Folders sections you cannot fill
  • Slash commands lead with the current chat's own commands, and a blocked composer control explains why instead of ignoring the tap
  • Drawers can be pulled closed from scrollable content, and a text selection wins over the sidebar swipe
  • Downgrading a plan explains what you would lose before it goes through
  • Coach marks stand down while a banner, the plan card or the mobile sidebar is up, then pick up where they left off
  • Fonts load faster now that the Latin faces are self-hosted, and loaders are refreshed across the app
  • Generated images render whole and can be downloaded
  • A Mermaid block keeps its reserved height after a failed render, and streaming keeps going when a comment follows the live turn
  • Right-to-left polish: mirrored text fields, inline code that stays in one piece, switches that stay in their track, and Arabic-script line heights
  • Safari joins Arabic letters while a reply streams, and the message clock stays on one line
  • Paying with a card already on file lands on the receipt instead of an empty page

v4.0.0

A Fresh New Lexika!

Our biggest update yet. Long-term memory evolves into Brains, the app is rebuilt on a local-first foundation for major speed gains, settings are fully revamped, and chat gets sharing, slash commands, shortcuts, a models gallery, and much more.

  • Introducing Brains: long-term memory reimagined — create multiple brains in your workspace and connect or disconnect each one to any conversation
  • Conversation sharing is here, and shared conversations can be forked by other users
  • New / slash commands for quick, keyboard-first access to all app actions right from the chat input
  • A dedicated models gallery page to explore every model in one place
  • Models now show their knowledge cutoff, the date they were added, and benchmark scores
  • Added prompt queues
  • Models can now react to your messages!
  • Keyboard shortcuts across the app, with a dedicated Shortcuts tab in Settings — including quick navigation through your last 9 conversations
  • A new island design for the sidebar that you can pick in Settings!
  • Manage your uploaded files in Settings — view, delete, and reuse already-uploaded attachments instead of uploading them again
  • New Knowledge Center, plus a fresh onboarding experience with helpful coach marks
  • Added a "Hide personal information" option to profile settings
  • Mermaid diagrams are now supported in chat
  • Date dividers in the chat UI so you always know when messages were sent
  • User message quick peeks on the right side of the chat for jumping through a conversation
  • Context compaction is now visible in the UI
  • Choose your own code block theme in Settings
  • New Download tab in Settings, and a reworked download page
  • Your favorite models are now stored in your account and synced across all devices
  • Data and state management rebuilt from the ground up with a local-first architecture
  • APIs now load up to 3x faster, with major UI performance improvements across the app
  • Settings fully revamped: preferences and settings are merged, new preference options were added, and mobile gets its own drawer design
  • The chat composer is completely refactored: drafts are now stable, and the character limit is increased to 20,480 with non-destructive over-limit handling and clear feedback
  • Each model now has its own set of supported reasoning effort options, with a brand-new effort selector UI
  • Chat rendering is completely improved: refreshed code blocks, more stable math formulas, and better image viewing
  • Council got updated presets and an improved model debates and answers pipeline
  • The automatic model router now includes new models, drops legacy ones, and picks smarter
  • Smarter context management, improved compaction logic, and better title generation for new conversations
  • Improved sign-in and sign-up UI
  • A refreshed home page and new-chat experience
  • Laboratory is redesigned into a more practical dialog and drawer
  • A smoother workspace invitation flow for guest users
  • Major UX improvements across inputs, conversation renaming, workspace settings, and sidebar folders
  • Improved landing page design and better UI stability when resizing
  • Fixed cross-tab sync issues for user sessions
  • Fixed a "workspace not found" issue when opening a conversation URL from a different workspace
  • Fixed conversation branching issues, including stale caches and incorrect timestamps after branching
  • Fixed support for text file attachments like Markdown, .py, and .js files

v3.2.3

Rendering, Authentication, and Reload Stability

This update improves LaTeX and math rendering, handles temporary authentication issues more gracefully, and prevents unexpected reloads after app updates.

  • LaTeX and math content now renders more reliably
  • Temporary authentication issues are now treated as recoverable so Lexika can keep the session moving
  • Chunk updates no longer cause unexpected automatic page reloads

v3.2.2

Workspace Invites, Chat Bubbles, and Model Access

This update fixes workspace invitation login behavior, chat bubble display, and model access after joining owner-funded workspaces.

  • Guest users opening a workspace invitation now see the login screen correctly
  • Messages from other workspace members now show with the correct chat bubble styling
  • New members in owner-funded workspaces can now use models immediately after joining, without refreshing the page
  • fix: prevent conversation status from incorrectly becoming IDLE during AI streaming

Follow us

Stay connected and get the latest updates from Lexika.