What’s New

cherry-studio

AI

cherry-studio release notes.

Latest v1.9.12 · by cherry-studioWebsiteCherryHQ/cherry-studio

Changelog

v2.0.0-rc.1

Pre-release
Added
  • Select knowledge bases from the input in agent composer
  • Sync skill catalog from disk so authored skills appear in agent skills
  • Warn before contextual model switches in agent chat
  • Add cost analytics and durable usage records in AI usage
Fixed
  • Allow selecting Markdown files from import hint in notes
  • Filter non-chat model choices in onboarding
  • Include sqlite-vec in portable Windows builds
  • Stabilize topic and session image export in message capture
  • Install both CPU architectures so x64 macOS builds ship sharp
  • Preserve migrated custom apps and icon sizing in mini-app
  • Preserve provider display names in provider models
  • Skip macOS process transform for workspace-visible windows in window registry
  • Improve trigram BM25 queries in knowledge search
  • Keep Bedrock auth modes switchable in provider settings
  • Show interrupted empty replies in message rendering
  • Preserve streams across tab switches in AI transport
  • Deduplicate custom size option in paintings
  • Correct user message presentation in quick assistant
  • Keep prompt fixed during image transforms in paintings
  • Stabilize regeneration and image export in message list
What's Changed

Full Changelog: https://github.com/CherryHQ/cherry-studio/compare/v2.0.0-beta.3...v2.0.0-rc.1

View originalPermalink

v2.0.0-beta.3

Pre-release
Added
  • Add managed CLI discovery and installation for agent-cli
  • Show client ID in developer mode in system-settings
  • Mask API key and Authorization Header with a reveal toggle in api-gateway
  • Offer a v1 download when a retried migration fails again
Changed
  • Disable model thinking for translation when supported
  • Pass model contextWindow as num_ctx in Ollama API requests
  • Rename Code Switch across locales
Fixed
  • Validate renderer URL origins in window navigation
  • Keep AskUserQuestion interactive in claude-code
  • Create channels inactive so credential-gated types can be added
  • Keep empty skills identity stable while the query loads
  • Refine global-search reveal and composer toolbar behavior
  • Stabilize streamed process history in message-renderer
  • Prompt when model is missing in agent-composer
  • Preserve state across approval overrides in composer
  • Preserve and render composer knowledge tools in chat
  • Mask generated API key in api-gateway
  • Support interactive QR authentication in agent-channels
  • Freeze chat topic titles after AI or manual naming
  • Apply registry updates to existing provider model configurations

[!WARNING] This Beta is not stable and is intended for testing only. Do not use it as a replacement for Cherry Studio v1. Data created by v2 may be cleared in later Beta releases or the final v2 release.

此 Beta 版本尚不稳定，仅供测试使用，请勿将其作为 Cherry Studio v1 的替代版本。v2 产生的数据可能会在后续 Beta 版本或 v2 正式版发布时被清除。

What's Changed
New Contributors

Full Changelog: https://github.com/CherryHQ/cherry-studio/compare/v2.0.0-beta.2...v2.0.0-beta.3

View originalPermalink

v2.0.0-beta.2

Pre-release
Added
  • Unify collection empty states with v1-style illustrations
  • Clearer stream-status indicators and awaiting-approval badge in session list
  • Enable transparent windows by default on macOS
  • Add copy action to span details in trace
  • Warmer runtime copy and rolling reasoning preview in agent message
  • Reference topics and sessions as @ mention context tokens in composer
  • Doubao and Bailian catalog enrichment with built-in web search
  • Bind knowledge bases with assistant-parity scoping in agent
  • Surface Claude SDK api_retry as ephemeral status in agent session
Changed
  • Adopt Chrome-style tab context menu and close behavior
Fixed
  • Accept nullable Bocha response fields in web search
  • Make read_file offset/limit schema Gemini-compatible in builtin tools
  • Delay conversation link preview cards in chat
  • Identify models in multi-model replies
  • Localize initial agent name for Chinese systems in cherry assistant
  • Constrain opacity slider height in selection assistant
  • Make local embeddings opt-in on creation in knowledge
  • Preserve models for recalled prompts in composer
  • Preserve artifact pane visuals across themes and tabs in right pane
  • Prevent command injection and path traversal in addModel

[!WARNING] This Beta is not stable and is intended for testing only. Do not use it as a replacement for Cherry Studio v1. Data created by v2 may be cleared in later Beta releases or the final v2 release.

此 Beta 版本尚不稳定，仅供测试使用，请勿将其作为 Cherry Studio v1 的替代版本。v2 产生的数据可能会在后续 Beta 版本或 v2 正式版发布时被清除。

What's Changed
New Contributors

Full Changelog: https://github.com/CherryHQ/cherry-studio/compare/v2.0.0-beta.1...v2.0.0-beta.2

View originalPermalink

v2.0.0-beta.1

Pre-release
Added
  • Upgraded interface with new interactions and visual design for chats, roles, projects, and tasks
  • Multi-window and split-view layouts for multitasking and switching between work and reviewing progress
  • Support for both OpenAI and Anthropic endpoints
  • Built-in Agent runtime core, reducing dependency installation and environment setup
  • Agents can read and maintain knowledge bases
  • Code tools upgraded to Code Mate with one-click model service configuration
  • Image generation dynamically shows parameters available for each model
Changed
  • Data foundation upgraded for improved responsiveness with extensive history
Fixed
  • Smoother scrolling in very long conversations
  • Large-folder imports no longer block the interface

[!WARNING] This Beta is not stable and is intended for testing only. Do not use it as a replacement for Cherry Studio v1. Data created by v2 may be cleared in later Beta releases or the final v2 release.

此 Beta 版本尚不稳定，仅供测试使用，请勿将其作为 Cherry Studio v1 的替代版本。v2 产生的数据可能会在后续 Beta 版本或 v2 正式版发布时被清除。

Cherry Studio 2.0.0-beta.1 - A New AI Workspace, from Chat to Getting Work Done

✨ New Interface, New Experience

  • Upgraded interactions and visual design make chats, roles, projects, and tasks clearer and easier to organize
  • Built for multitasking, with multi-window and split-view layouts for switching between work and reviewing progress and deliverables

🤖 More Powerful Agents

  • Supports both OpenAI and Anthropic endpoints, reuses Claude and Codex subscriptions, and works with more models
  • The core Agent runtime is built in, greatly reducing dependency installation and environment setup
  • Complete research, documents, presentations, data analysis, and coding faster, with one-click previews of results

🛠️ More Tool Upgrades

  • Agents can read and maintain knowledge bases, helping your knowledge grow continuously like a Wiki
  • Code tools are upgraded to Code Mate, with one-click model service configuration for Claude Code, Codex, and more
  • Image generation dynamically shows the parameters available for each model, making results easier to control

⚡ Performance

  • A fully upgraded data foundation stays responsive even with extensive history
  • Smoother scrolling in very long conversations, while large-folder imports no longer block the interface
View originalPermalink

v1.9.12

Added
  • Add Kimi K2.7 Code support
Changed
  • Redirect provider links to inferera and follow configured base URL in aihubmix
Fixed
  • Forward DeepSeek V4+ reasoning_effort for openai-compatible providers
  • Support Claude Fable series and Opus 5 in capability detection
  • Add Anthropic native /v1/models fetcher
  • Prevent AccordionTrigger toggle when clicking approval buttons in collapsed group header
  • Use anthropicApiHost for NewAPI provider's anthropic endpoint
  • Inject Cherry Studio identity headers into agent mode requests
  • Support downloading SVG data URL images blocked by CSP
  • Restore leading dot in ext for saveBase64Image and savePastedImage
  • Correct MiniMax-M3 thinking control
  • Surface abnormal stream finish reasons as error chunks
  • Preserve full URLs and parse Highlights in ExaMCP results
  • Recover claude-code install when native binary is missing
  • Derive Gemini baseURL from configured baseURL in aihubmix
  • Use first configured key for claude code agents
  • Use fs.statfs for disk info to stop powershell process leak
  • Hide CherryIN manual model add button
What's Changed

Full Changelog: https://github.com/CherryHQ/cherry-studio/compare/v1.9.11...v1.9.12

View originalPermalink

v1.9.11

Added
  • Add MiniMax M3 thinking process support
  • Support Anthropic Opus 4.8 adaptive thinking
Changed
  • Use openPath instead of openExternal for HTML artifacts external browser
Fixed
  • Encode gemini model list API key
  • Infer custom provider model groups
  • Clean up agent session messages on delete
  • Fix oversized text in Gemini mini-program due to Electron UA detection
  • Avoid embedding dimensions request parameter
  • Bind CherryIN OAuth flow to sender
  • Upgrade mermaid to 11.15.0 to fix CJK parsing error in flowcharts
  • Fix agent task completed but task list still shows unchecked items
Removed
  • Prune legacy MiniMax models, keeping only M2.7 and M3
What's Changed
New Contributors

Full Changelog: https://github.com/CherryHQ/cherry-studio/compare/v1.9.9...v1.9.11

View originalPermalink

v1.9.9

Added
  • Add MiniMax M3 to the system model registry
Fixed
  • Guard OV OCR availability CPU detection
  • Respect topic naming setting for agent sessions
What's Changed
New Contributors

Full Changelog: https://github.com/CherryHQ/cherry-studio/compare/v1.9.8...v1.9.9

View originalPermalink

v1.9.7

Added
  • Support StepFun as Anthropic-compatible provider
Changed
  • Change Gemini Safety Settings to BLOCK_NONE
Fixed
  • Clarify OpenClaw external install prompt
  • Support Grok 4.3 reasoning effort in xAI responses
  • Keep light theme tokens readable in code view
  • Apply [1m] context suffix for DeepSeek V4 Flash and MiMo V2.5+ in claudecode
  • Pass custom headers to claude code in agents
  • Launch opencode from package-local executable
  • Align Gemini 3.x UI and sampling handling
  • Add missing key prop to SendMessageButton in InputbarCore
  • Correct provider ID for AIHubMix reasoning effort
  • Support capabilities for grok-build-0.1
  • Exclude Qwen max series from vision model detection
  • Respect enableDataCollection setting in renderer for analytics
  • Do not treat Gemini model via non-Gemini provider as native PDF
  • Align ExaMCP response parser with server field names in websearch
  • Add CherryIN DeepSeek 1m suffix
  • Keep note toolbar static at top during scroll
  • Refresh cached miniapp URLs
What's Changed
New Contributors

Full Changelog: https://github.com/CherryHQ/cherry-studio/compare/v1.9.6...v1.9.7

View originalPermalink

v1.9.6

Fixed
  • Preserve HTTP URLs in knowledge base documents
  • Thread idle timeout handle to stream chunk adapter
  • Improve GitCode sync reliability
  • Convert assistant image blocks to base64 for multi-turn editing
  • Invalidate provider cache on settings changes
  • Recognize ctrl by uniKey value
  • Prevent crash when model.provider not found
  • Align provider model list actions
  • Use task name as session name for cron task sessions
  • Resolve ENOENT in OpenMinerU preprocessing caused by MinerU ZIP structure change
  • Clean up OAuth tokens when deleting MCP server
  • Exclude deepseek-r1 models from function_calling capability
  • Auto-expand MCP tool approval card so action buttons are immediately visible
  • Add environment variable for claude code provider management
  • Preserve updatedAt timestamps when creating new agents
  • Qiniu PDF fallback for GPT-5.4
  • Handle Codex CLI reserved provider IDs
What's Changed
New Contributors

Full Changelog: https://github.com/CherryHQ/cherry-studio/compare/v1.9.5...v1.9.6

View originalPermalink