Bionic 1.2.1

November 16, 2026

1.2.1 - Release Notes

  • LM Link remote chats and projects on other Bionic and llmster devices.
  • Device and folder selection for remote projects and working folders, with a compact folder picker.
  • Fast file transfers between remote agent devices.
  • Remote machines settings with an “Allow remote agent control” switch.
  • Remote agents onboarding with a shortcut to link devices.
  • Remote machine sidebar with device management and remembered chat expansion.
  • Clearer connection status when reopening remote chats.
  • Clearer remote chat errors, with retry options for connection problems.
  • Sign-in prompts in remote chats after logout.
  • Other devices hidden when LM Link is turned off.
  • Clearer settings navigation.
  • Consistent sidebar header controls when collapsed.
  • Last-used model fallback for new assistant-created chats.
  • Fixed missing model picker in chats without a selected model.
  • Fixed remote devices lingering after logout.
  • Fixed unavailable remote devices blocking new local chats.
  • Fixed misleading update prompts for disconnected LM Link devices.
  • Fixed model loading failures after switching to the ARM64 build on Windows.
  • Fixed AGENTS.md instructions being lost after conversation compaction.
  • Fixed missing or misnamed tool arguments with compatible external API models.
  • Fixed missing spacing around the Bionic download link in Chinese, Japanese, and Korean.
  • Clearer guidance to install the ARM64 build on Windows ARM PCs.
  • Clearer project creation dialog, with labeled fields and a close button.
  • More complete translations for Remote settings, devices, onboarding, and scheduling.
  • llama.cpp 2.76.0 extension packs.
  • oMLX 0.4.0 extension packs.
  • Opt-in beta API for launching assistants headlessly and streaming their progress.
  • Fixed blurry image previews on high-density displays.
  • Fixed model downloads stuck on “Resolving model files” when remote download permission is missing.
  • llama.cpp 2.74.0 extension packs.
  • Settings to allow assistants to read or operate other sessions without permission prompts.
  • Conversation timestamps in chats, with separators after long gaps and day changes.
  • Support returning images in MCP tool call results.
  • Tighter input and search field spacing.
  • Fixed incorrect GPU offload settings for GGUF models with large metadata headers.
  • /archive command for chats.
  • Local / Cloud / All model filters, with the last selection remembered.
  • Consistent model picker sizing, with long names scrolling on hover.
  • Fixed flickering Free badges and model list jumps during offer refreshes.
  • Projects in Cmd+P, opening in dedicated project windows.
  • “Open in Project Window” in workspace chat and Topic tab menus.
  • Experimental support for external API models.
  • Consistent project and session dialogs.
  • Warning when Minimum AutoFit context length is below the default.
  • Fixed missing onboarding on fresh installs.
  • Fixed startup crashes in public builds.
  • Fixed hidden read-aloud settings when voice transcription is unavailable or disabled.
  • Fixed memory growth while chats are idle.
  • More complete translations across supported languages.
  • Consistent checkboxes in task lists and settings.
  • Tighter Markdown heading spacing.
  • Cleaner sidebar styling.
  • Clearer guidance when MLX models run out of context space.
  • oMLX 0.3.0 for Apple Silicon Macs, with clearer Apple M5+ labeling.