Changelog 👾

Build 3

• Markdown preview and slash-command spacing polish.
• Fixed LM Link compatibility handling for remote context-length lookups.
• Fixed app icon sizing on Windows.

Build 2

• Document tools for the coding agent.
• LM Link remote model configuration and loaded-instance status in the Model Library.
• Voice model auto-loading tuned to available Mac memory.
• Safer handling of destructive Git commands.
• Fixed Gemma 4 native tool-use detection.

Build 1

• Runtime downloads hidden by default for a cleaner downloads list.
• Clearer document-processing and sandbox transcript labels.
• Settings toggle and Connected Apps UI polish.
• llama.cpp 2.26.0 extension packs.

Build 3

• UX improvement in model picker
• Refresh cloud model list when refreshing billing information in Settings
• Nicer UX showing when skills are used
• Fix Settings title bar not draggable
• MCP OAuth fixes, including public discovery, reauthentication, and custom scopes.
• Fix issue with model model selection not persisting after project creation
• Upon project creation, create a new tab in that project

Build 2

• Fixed first-run onboarding for Bionic 1.0.x.

Build 1

• Exploration agents setting in Chat settings.
• Better Hugging Face model card rendering.
• Less project search loading flicker.
• Fixed MLX model load cancellation.

Build 1

• Model picker and voice model download layout polish.
• Fixed duplicate workspace windows during startup.
• Fixed Windows app icon.
• Fixed duplicate LaTeX rendering.
• Fixed downloads button placement on Windows.
• Fixed turn-summary diffs across line endings.

Build 1

• LM Studio Bionic initial release