Esta é uma versão antiga. A versão atual é 1.2.0+2. Bionic 1.2.0+2 · Lançamento:
Arquivos de instalação 1.1.0+11
| Arquitetura | Tipo | Escopo | Tamanho do arquivo | Baixar | SHA256 |
|---|---|---|---|---|---|
| 64 bits | EXE (NSIS) | — | — | bionic-installers.lmstudio.ai | VirusTotal |
O link e o SHA256 vêm do manifesto comunitário do Gerenciador de Pacotes do Windows. Compare o hash do arquivo baixado antes de executá-lo.
Notas da versão
- Project-less chats
- Auto Review for shell commands
- Simpler project creation: unify code and work project types, with optional coding abilities per session
- Support for Images in tool call results in the /v1/responses, /v1/chat/completions, and /v1/messages APIs
- Git agent instructions adjustment
- llama.cpp 2.31.0 extension packs
- Fewer approval prompts for shell commands implied by your request.
- Fixed right panel headers blocking window dragging.
- Localized standard reasoning-level labels for LM Link models.
- More consistent Project Files folder controls and previews.
- Security hardening
- Word document comments and tracked changes omit timestamps unless requested.
- Project code-permission guidance in the New Project modal.
- Fewer approval prompts for safe read-only GitHub queries.
- Fixed LM Link connections to newly discovered peers.
- Vision subagents: set a local or cloud model to be process images
- Fewer approval prompts for additional safe shell commands.
- Persistent shell approval preferences and custom Auto Review instructions.
- Fewer approval prompts for safe LMS and ESLint commands.
- Cross-project session moves.
- Fixed the command approval tooltip remaining visible after opening the menu.
- Consistent sizing and spacing for composer controls.
- Fixed transcription stop button colors.
- No-project session section in the sidebar
- Composer controls for disabled, manual review, auto review, and allow-all shell modes.
- Vision subagents for letting text-only models understand images.
- Local model loading and prompt-processing progress in transcripts.
- MLX prompt disk cache controls in advanced model settings.
- Muse Glimmer tool calling for MLX models.
- Fixed composer chips disappearing across draft edits.
- Fixed the GPU settings pop-out.
- Fixed some UI lags while the agent is working
- Fixed shell commands failing with zsh configurations that replace default aliases.
- llama.cpp 2.30.0 extension packs.
- Fixed transcript scroll position jumping during tool interactions.
- UI performance improvements
- More reliable LM Link connections during high-volume activity.
- Fixed Office document previews failing after a cancelled conversion.
- 2–2.75× faster prompt processing on M5 Macs.
- Fewer approval prompts for safe read-only shell and GitHub commands.
- Syntax highlighting for more languages in project file previews.
- Horizontal scrollbars for wide Markdown tables now appear on hover.
