Version History

Version 4.4 Tycho

DEVONthink 4.4 turns the AI assistant into a true co-worker. A new AI library lets you create, store, and sync roles, prompts, and skills that can trigger your smart rules, batch processes, and scripts on demand. In addition, the Chat assistant can use DEVONthink's MCP tools directly and improved prompt caching makes multi-step tasks noticeably faster and much cheaper. Of course, we have also updated to the latest models of all major AI providers.

Markdown takes the second spotlight with WYSIWYG support for live tables and MathJax equations, a new syntax colorizer, custom color themes, and direct export to LaTeX or typesetting as PDF via pdflatex. Web clipping is also faster and more accurate, using a new, local Markdownifier and Readability engine. The new Reader view lets you read web content distraction-free.

Finally, encrypted and revision-safe databases use modern APFS images on macOS 26.0 Tahoe and later, and the help viewer's chat assistant is available for free in all editions.

• DEVONthink adds two new export commands to the File > Export menu for exporting Markdown documents to native LaTeX or PDF. LaTeX export resolves WikiLinks and transclusions, carries over title, author, and date metadata, and copies referenced images alongside the exported document; PDF export requires a local installation of MacTeX or the smaller BasicTeX, and merges multiple selected Markdown documents into a single PDF.
• DEVONthink adds a new Reader view that displays a clean, sanitized read of web content like bookmarks, HTML documents, and web archives that contain a detectable article. Open it via View > Document Display > Reader, the Reader button in the Navigation Bar, or ⌃︎⌘︎A, then capture the page via the Tools > Capture menu, a context-menu command, or by printing.
• DEVONthink uses a new, local engine for converting web content to Markdown. Beyond speed and native compatibility, it also improves navigation, advertising, and boilerplate removal; metadata, embedded media, and paywall detection; forum thread and authorship detection; callouts and footnotes; and MathML-to-LaTeX conversion.
• DEVONthink adds a Color Theme popup to the Files > Markdown settings, letting you choose a theme for both the source and preview modes of Markdown documents. The Automatic option uses the Markdown font and plain text colors chosen in the Editing > Format settings.
• DEVONthink adds a Date Attributes Modified criterion to the predicate editor for smart groups, smart rules, and advanced toolbar searches. The search prefix is attributesModified or attributesModification and honors the same date-based prefix operators as other date criteria; in the higher editions, the MCP server and Chat assistant support this property too.
• A new read-only AppleScript property, has text layer, lets you quickly identify PDF documents that need OCR.
• DEVONthink brings a new mark record AppleScript command that adds annotations (highlights, underlines, and strikethroughs) to found text within PDFs with text layer, rich text, Markdown, and HTML-based documents like formatted notes. The MCP server supports this too.
• DEVONthink adds new application scripts for use with MailMaven: Add attachments to DEVONthink, Add message(s) to DEVONthink, and Add message(s) & attachments to DEVONthink. Importing attachments requires full disk access.

• Pro — DEVONthink adds an extensive new AI > Library settings pane: a repository of roles, prompts, and skills you can use in the Chat assistant and in smart actions. It ships with ten built-in, modifiable functions, each with its own settings and a keyword (slash command) for quick access, and you can create your own manually or with AI assistance. Skills can call scripts (including AI-scripted ones) and trigger existing automations like smart rules, batch processing, and Script with Input/Output actions. Your library syncs via iCloud across your Macs and can also be exported as a package to share. Skills require AI models capable of using MCP tools.
• Pro — DEVONthink adds an Allow MCP tools option to the AI > Chat settings and the Chat assistant, giving the AI access to DEVONthink's MCP tools while you chat. The tools honor your other assistant settings and require a model that supports tool calls with a context window of at least 64,000 tokens (currently ChatGPT, Claude, Mistral, Gemini, and Qwen 3.x and Muse Spark via OpenRouter). DEVONthink caches requests to minimize token spending, though MCP calls still take more time and tokens than a plain chat.
• Pro — DEVONthink adds a Copy JSON button to the AI > MCP > Install Connectors settings to copy the stdio configuration for use in third-party applications, and a similar Copy Server JSON Config command in the MCP Menu Extra for the HTTP access configuration.
• Pro — Tooling-capable AI models can retrieve a PDF's page count in the Chat assistant, and the AI tool that extracts record contents accepts a new pages parameter to target specific pages or a page range. Available to the MCP server, the Chat assistant, AI smart actions, and the get chat response for message AppleScript command, this lets you process large PDF documents in smaller chunks and stay under the context window while reducing cost and processing time.
• Pro — A new AppleScript command, perform chat skill, runs an existing skill and returns its results — handy for running skills on a schedule via smart rules. Skills that require confirmation can't be run via scripting.
• Pro — A new tool retrieves a PDF document's table of contents in the Chat assistant, the MCP server, smart actions, and the get chat response for message AppleScript command.
• Pro — DEVONthink supports Meta's Muse Spark 1.2 via OpenRouter.
• Pro — DEVONthink supports macOS 27's fm serve server as an OpenAI-compatible provider.
• Pro — DEVONthink supports Google's Nano Banana 2 Lite for fast image generation, though HD quality isn't available.
• Pro — DEVONthink adds prompt caching for Mistral AI and Claude models served via OpenRouter.

• We significantly improved the appearance and reliability of the DEVONthink Markdown parser and renderer. Beyond the new color theme option, the WYSIWYG source now more closely matches the rendered preview, and the WYSIWYG view supports live-edited tables and MathJax equations, not just images and links. DEVONthink also natively renders GitHub-style emojis.
• You can now copy multiple selected ranges in the Markdown WYSIWYG editor.
• Several inspectors use the new parser for consistency, including Info pane > Properties, Content > Table of Contents, Document > Annotations, Attachments, and Links.
• DEVONthink produces more naturally readable mathematical output when converting Markdown documents with equations to plain text.
• Markdown and HTML documents preview more responsively in Side-by-Side view.
• DEVONthink wraps long lines in code blocks within the Markdown preview and Reader view, instead of letting them run past the visible width of the pane.
• The default stylesheet set in the Files > Markdown > Style Sheet setting no longer overrides DEVONthink's internal stylesheet, retaining more of the built-in styling when you edit just a few CSS elements.
• DEVONthink clips web content directly from the loaded WYSIWYG page, more reliably stripping unwanted elements like cookie banners while retaining important elements such as images and scripts, including on sites like YouTube or X.
• When clipping, DEVONthink uses browser-supplied HTML (e.g. by bookmarklet, extension or via AppleScript) more reliably by keeping pages from refreshing or forwarding themselves.
• If you've viewed and logged into a paywalled or authenticated site in DEVONthink, the app can reuse the stored cookies when you clip from that site externally, e.g., via our browser extension, increasing the chance of a successful capture.
• DEVONthink better converts inline code in web content to Markdown.
• DEVONthink lets you toggle the Clutter-free layout when clipping web content as Markdown via the Sorter and the create Markdown from AppleScript command.
• When composing a Markdown note in the Sorter, DEVONthink imports images you drag and drop into it, if the Files > Markdown > Import images to group option is enabled.
• Text annotations added to rotated pages of PDF documents are horizontal now, following the current rotation of the page instead of its original orientation. Their default size, their automatic resizing to fit the text, and their resize handles follow the rotation too.
• In a sheet created with the Tools > Metadata Overview command, DEVONthink defines the Item Link column as a crosslink instead of a URL.
• You can select and copy content in cells within read-only sheets, e.g., locked or revision-proof databases.
• DEVONthink creates encrypted and revision-proof databases as APFS disk images on macOS Tahoe and later instead of using the older HFS+ format. Note that you cannot open databases using the new APFS encryption on older versions of macOS.
• DEVONthink unmounts encrypted and revision-proof databases more reliably.
• DEVONthink supports the matches (:) operator, with quoted or unquoted text, when searching custom metadata, e.g., mdstatus:"To Do".
• Going to a page, line or row of a sheet no longer depends on the system's number formatting, both 1018, 1 018, and 1,018 work now.
• Multiple selected items with thumbnails display as a fanned stack in the view/edit pane.
• The Info filter supports filtering by Date Attributes Modified.
• We made many minor cosmetic tweaks and fixes to the interface, touching inspectors, the splash screen, the License panel, etc.
• The Databases Finder folder has a new icon on macOS Tahoe and later.
• We updated the layout of the Help viewer and the help pages, including a reordered sidebar.
• DEVONthink makes the Help assistant available for free in all editions and during the trial period. We are providing this functionality as a courtesy, but note that you can no longer change its AI provider.
• DEVONthink retrieves ISBN numbers more reliably via AppleScript's international standard book number property and smart action placeholders.
• We renamed the AppleScript property attributes change date to attributes modification date, keeping the old name as a synonym for compatibility.
• The replicate record to AppleScript command supports replicating to an RSS feed.
• The Export > Listing script runs at least six times faster.
• We improved language detection for documents, making it faster and more accurate.
• DEVONthink processes and indexes HTML content more effectively, removing unnecessary information to improve searches and conversions.
• DEVONthink clears smart group and smart rule caches immediately when you close a database.

• Pro — Imprinted documents achieve better compression when saved.
• Pro — The Chat assistant and MCP server better handle Markdown documents with empty sections.
• Pro — You can file documents and groups created via the Chat assistant by providing a name or path.
• Pro — In the AI > Chat settings, DEVONthink updates per-model usage costs after each completed request using current rates, for more precise totals. Costs also reflect the discounted rates for cached data where applicable, and show an estimated percentage of cached input.
• Pro — All the AI functions, e.g., the Chat assistant, smart actions, scripting commands, AI popovers, etc., redact your name and phone number, and the phone numbers of your other contacts. The MCP server optionally does this as well.
• Pro — Better caching and response filtering reduce costs and increase accuracy in the Chat assistant, AI smart actions, and the get chat response for message AppleScript command.
• Pro — We improved error logging and handling for the Chat and Help assistants.
• Pro — Local MCP server connections no longer require a TLS certificate, though one is still recommended.
• Pro — In the AI > MCP settings, Install Connectors > Codex Desktop & CLI supports the latest ChatGPT application.
• Pro — The MCP server can get color label names, more flexibly access the Global Inbox, and better handle UUIDs.
• Pro — DEVONthink's MCP server supports the latest version of MCP as well as previous versions. This improves compliance and reliability in connection handling and also prepares the server component for future MCP-compatible applications.
• Pro — All MCP tools declare the scope of their actions via annotations. These denote whether an action can access the internet, modify or overwrite data, or is read-only. AI clients that support these annotations, e.g., Claude Code, use them to determine if user confirmation is required.
• Pro — We updated Claude Opus and Sonnet to version 5, with a knowledge cutoff of 2026-07-01; Sonnet has a context window of one million tokens and 128,000 output tokens. Pricing hasn't changed, though the new tokenizer uses up to 30% more tokens per request. Claude's prompt caching is much more efficient, making multi-step tasks noticeably faster and reducing token costs.
• Pro — DEVONthink supports the latest OpenAI models, GPT 5.6 Sol, Terra, and Luna, each with a one-million-token context window, 128,000 output tokens, and knowledge current to February 16, 2026. The Script assistant uses 5.6 Sol.
• Pro — We updated Gemini Flash to 3.7 and Flash Lite to 3.1, both accessible via API and OpenRouter, with knowledge current to 2026-07-01.
• Pro — DEVONthink uses the OpenAI responses API and Google's interactions API, supporting the latest models and features, including image tool call results.
• Pro — DEVONthink offers improved support for running GLM 5.x and muse-glimmer models locally.
• Pro — DEVONthink supports context windows of up to one million tokens for Ollama, LM Studio, and OpenAI-compatible providers, up from 128,000.
• Pro — Local AI, Ollama, OpenAI-compatible, and Mistral models support longer tool call cycles.
• Pro — DEVONthink offers OpenAI's GPT 5.5 (Pro) via the API.
• The Script assistant supports GPT 5.3 Codex. We updated the OpenRouter models used by the assistant: Claude Opus and Sonnet 5, Gemini 3.7 Flash, Grok 4.5, and Kimi 2.7 Code.
• Pro — MCP tools and AI skills support Qwen 3.x and Muse Spark models via OpenRouter.
• Pro — Transcriptions without timestamps done via OpenAI use GPT-Transcribe instead of GPT-4o.
• Pro — The Script assistant and AI skills produce better AI-generated scripts.

• Selecting text in the Chat assistant could sometimes crash on macOS 27 Golden Gate. Fixed.
• Viewing or editing Markdown documents containing lists could sometimes crash. Fixed.
• Markdown documents converted to HTML incorrectly contained a second title header. Fixed.
• MathJax in Markdown didn't always render properly. Fixed.
• A Markdown document with unclosed YAML front matter at the top of the document displayed and converted as if it had no content. Front matter now ends at the first blank line, similar to MultiMarkdown. Fixed.
• In Markdown, bold or italic text followed by a superscript character wasn't formatted correctly. Fixed.
• Line items spanning multiple lines in Markdown documents' to-do lists incorrectly showed the text on the line under the checkbox. Fixed.
• With Syntax highlighting enabled in the Files > Markdown settings, callout titles in the Markdown source didn't display in bold and tinted colors like in the rendered view. Fixed.
• Saving edits to Markdown and HTML documents in side-by-side view forced a full reload of the preview, causing flickering and rescrolling. Fixed.
• When text was selected in the Markdown source while using the Side-by-Side view, the Edit > Set Name As command was disabled. Fixed.
• The Smart quotes and dashes option in the Files > Markdown settings got disabled after restarting DEVONthink. Fixed.
• Saving edits to formatted notes could reprocess styles and images, leading to image scaling issues. Fixed.
• Clipping HTML pages with malformed markup, legacy multimedia content, or self-closing tags could cause the content to be lost. Fixed.
• Clipping to formatted notes could cause random issues. Fixed.
• PDF annotations could get lost or damaged on macOS 27 Golden Gate when editing them or when pages were copied, e.g., by splitting, merging, or converting documents. Fixed.
• Searches beginning with NOT incorrectly counted the first term twice, inflating the item's ranking. Fixed.
• Searches using the NOT AFTER operator didn't match documents whose search term was the last word of the text. Fixed.
• Exact phrase searches, e.g., "a b c", could return similar but non-matching documents, e.g., one containing "a b b c". Also, if a document contained repeated words, e.g., the cat the cat the dog, a quoted search for "the cat the dog" didn't match that document. Fixed.
• Sorting very large groups or smart groups by URL could sometimes cause a crash in recent versions of macOS. Fixed.
• Windows opened via the context menu in the Navigate sidebar or item list would open in the background on macOS 26 Tahoe and later. Fixed.
• The toolbar search field in the Help viewer wasn't vertically centered on macOS 27 Golden Gate. Fixed.
• The Navigation Bar was sized incorrectly when previewing Markdown documents, HTML pages, or web archives on macOS 27 Golden Gate. Fixed.
• When inserting a back link or quote in an annotation file, Asian characters in Markdown section names were converted to Latin. Fixed.
• Documents with text containing the name or alias of another document weren't shown in the Document > Mentions inspector if the first word of the name was repeated, e.g., skill skill prompt didn't match a document named skill prompt. Fixed.
• When the maximum number of versions per document was reached, the next save would incorrectly remove the most recent version instead of the oldest. Fixed.
• You couldn't drag and drop multiple emails from Apple Mail into DEVONthink on macOS 27 Golden Gate. Fixed.
• When trashing a replicant matched by a local smart group, the menu and context menu command was "Move to Trash" instead of Move All Replicants to Trash. Fixed.
• The Scan Name and Scan Text smart actions could cause an exception when processing a document with no text, e.g., a bookmark. Fixed.
• Scheduled smart rules that hadn't run due to having no matched items could cause high CPU usage. Fixed.
• An alias or symbolic link in the Scripts or Templates menus that pointed to an enclosing folder could make DEVONthink freeze and finally crash. Fixed.

• Pro — When browsing certain sites, too many items were added to the Download Manager. Fixed.
• Pro — If the MCP server failed to launch on startup, the alert always showed the default 8420 port instead of the one configured in the MCP settings. Fixed.
• Pro — Summarizing or tagging large documents with Apple Intelligence could produce frequent "Exceeded model context window size" errors. Fixed.
• Pro — Gemini didn't work reliably via OpenRouter. Fixed.

• Server — In the web interface, dates changed in the Data inspector were not saved. Fixed.
• Server — Some Markdown documents in search results couldn't be edited. Fixed.
• Server — The Generic and Data inspectors were disabled when a search result in the web interface was selected. Fixed.

• We removed MultiMarkdown support and its renderer popup from the Files > Markdown settings.
• We removed the Clutter-free layout option from the Tools > Capture menu, as it's no longer needed with in-application clipping.
• We removed the obsolete iData file plugin (e.g., .idata, .id3).
• We removed a workaround of the List and Icons views in widescreen preview mode on the latest versions of macOS 26.x so that the scrollbar is not always visible anymore but honors the system's setting.

• Pro — We removed the outdated O3 and O4 Mini models from OpenAI access and OpenRouter.
• Pro — We removed Magistral models from Mistral and OpenRouter access.
• Pro — We removed Meta Llama 4 Maverick and Scout, and DeepSeek R1, from OpenRouter.
• Pro — We replaced several OpenRouter models: GLM 5.1 with GLM 5.3, DeepSeek V4 Flash 0423 with DeepSeek V4 Flash 0731, DeepSeek V4 Pro with DeepSeek V4 Pro 0813, Qwen 3.6:27B with Qwen 3.8:27B, Qwen 3.6 Flash with Qwen 3.7 Flash, Qwen 3.7 Max with Qwen 3.8 Max, and Grok 4.3 with Grok 4.6.
• Pro — We removed the Role popup from the AI Chat settings; specify roles in context instead, e.g., via a skill.

Version 4.3.2

This maintenance release adds support for Anthropic's latest frontier model, Claude Fable 5, as well as image diffusion models via Ollama. We optimized the search tool in the MCP server and added a built-in connector for LM Studio to access it. Additionally, DEVONthink can redact many national identification numbers before AI use. For research automation, four new AppleScript commands let you locate and apply bibliographic metadata, and in the web interface, you can now print PDF documents you're viewing. We also fixed compatibility with macOS 27 Golden Gate Developer Beta.

• Added four new research-based AppleScript commands: resolve DOI metadata, resolve book metadata, download PDF from DOI, and search CrossRef. These complement the functions of the MCP server in the higher editions of DEVONthink.
• Added AppleScript synonyms: DOI for digital object identifier and ISBN for international standard book number.

• Pro — Added support for Claude Fable 5, via Anthropic and OpenRouter. Be aware that this is a much more expensive model to use. Also note that at the time of the release of this version, Anthropic has suspended access to Claude Fable 5 based on a US government directive. It may or may not become available in the future.
• Pro — Added LM Studio to the AI > MCP > Install Connectors settings.

• Improved the overall reliability of the DEVONthink Markdown parser.

• Pro — Updated the AI > Mixture of Agents - Images template.
• Pro — DEVONthink redacts regional identification numbers, e.g., European identifiers, before passing them to external AI via the Chat inspector, AI-related smart actions and scripts, the Summarize and Transform popovers, and optionally the MCP server.
• Pro — The Chat assistant and smart actions no longer redact phone numbers, consistent with the MCP server's behavior.
• Pro — You can manually edit the MCP server's config.json file and retain your changes, unless you toggle the Redact sensitive content option in AI > MCP settings.
• Pro — In the Generate Image window or AI > Image Generation settings, you can now choose image diffusion models provided by Ollama, e.g., z-image-turbo. These models are also available via the model parameter of the download image for prompt AppleScript command and the Mixture of Agents - Images template.
• Pro — Added optional from_uuid parameter to the move_record MCP tool to move only replicants from the specified source group. Ignored in case of cross-database moves.
• Pro — Optimized and refined the search tool and the overall performance and reliability of the MCP server.
• Pro — The default context window for LM Studio is now 8,192 tokens.
• Pro — Added support for Nvidia Nemotron 3 Super/Ultra and MiniMax M3 to OpenRouter.
• Pro — More reliable get chat response for message … record … as "JSON" AppleScript command.
• Pro — The Download Bibliographic Metadata smart rule now uses the new AppleScript commands.

• Server — In the web interface, you can now print PDF documents for which you have download rights.

• DEVONthink crashed on launch on the macOS 27 Golden Gate Developer Beta. Fixed.
• In Markdown documents, bold or italic formatting applied to selected text already containing formatting didn't properly apply to the whole selection. Fixed.
• Double-clicking a PDF document in the Reading List didn't always open the document on the correct page. Fixed.
• Viewing or annotating a PDF document on the macOS 27 Golden Gate Developer Beta caused exceptions. Fixed.
• The see-also score for images was usually too high, e.g., for scanned documents. Fixed.
• Detached popovers weren't always restored properly in the latest versions of macOS. Fixed.
• The DEVONthink widget extension and the MCP server (in the higher editions) contained a file that could cause Time Machine to report an error. Fixed.
• The AppleScript search command could return results not sorted by score. Fixed.
• Window resizing was slow after clicking the Update Now button which is shown if the license is only valid for updates for less than 30 days. Also, after extending the application license, the Update Now notification could incorrectly continue to appear. Fixed.

• Pro — Removed the prompt strength parameter from the download image for prompt AppleScript command.
• Pro — Removed the outdated Stable Diffusion 3.5 models as well as the deprecated GPT-Image-1-Mini model.

Version 4.3.1

This maintenance release further refines the new DEVONthink Markdown parser with options like better image styling and captions, improved inline footnotes, and some fixes. This release also improves the MCP server, including safer handling of custom metadata, setting item aliases, opening documents externally, and using OpenLibrary as a fallback for Google Books. Additionally, we updated some image-generation AI models and now support Claude Opus 4.8.

• Support for headerless tables in DEVONthink Markdown.

• Pro — A new optional parameter for the open_record MCP tool allows you to open documents in an external application. This is disabled by default.
• Pro — An exclude from chat AppleScript property that we added to the database class lets you explicitly hide databases from AI access.

• DEVONthink's Markdown parser better supports styling directly in image links, including CSS units for scaling.
• DEVONthink renders alt text in an image link as an image caption and supports standard Markdown styling or linking, e.g., ![*Polar Bear*](/Animals/polarbear.png).
• Inline footnotes in DEVONthink Markdown preserve embedded links.
• Anchors in Markdown section headings are compatible with the MultiMarkdown implementation.
• Improved overall reliability of our Markdown parser and renderer.
• Updated Finder icons for all DEVONthink document types, e.g., .dtBase2, .dtSmartRule, etc., on macOS Tahoe.

• Pro — Updated AI > Mixture of Agents - Images template.
• Pro — You can now change an item's alias via MCP.
• Pro — Replicating items via MCP reports the destination of each item.
• Pro — The Chat assistant, the Chat - Query and Chat - Continue if smart actions, the get chat response for message and display chat dialog AppleScript commands, as well as the summarization and transformation chat popovers now also redact IBANs. The MCP server redacts IBANs if redaction is enabled.
• Pro — The MCP tool to resolve Google Books metadata now uses OpenLibrary as a fallback, as Google Books requires an API key to search.
• Pro — Boolean custom metadata is set more reliably via MCP.
• Pro — MCP tools in general are more reliable, including error responses and handling missing UUIDs for groups. We have improved the communication with the MCP server too.
• Pro — On macOS Tahoe, the Apple Speech Analyzer more reliably transcribes audio and video when an invalid language is specified via MCP or AppleScript.
• Pro — For the download image for prompt AppleScript command, we updated the name of some AI engines and also the description of some parameters. However, synonyms ensure older scripts are still compatible.
• Pro — We updated Claude Opus to version 4.8, via Anthropic and OpenRouter.
• Pro — We replaced GPT-Image-1.5 with GPT-Image-2. Note that "HD" quality produces larger images.
• Pro — Nano Banana 2 uses the commonly available version gemini-3.1-flash-image model.
• Pro — Editing images with Flux Pro or Max retains the original's image size, if possible.
• Pro — We updated the Recraft image-generation model, but note that it no longer supports styles.
• Pro — DEVONthink sanitizes Gemma 4 responses via Ollama.

• Server — Added localization to the See Also & Classify inspector in the web interface.

• The Markdown parser incorrectly detected an indented horizontal line with spaced characters, e.g., * * *, following a list in Markdown as a heading. Fixed.
• Callouts and blockquotes beginning with lines similar to MultiMarkdown metadata didn't parse and render correctly. Fixed.
• In Markdown documents, emboldened text following punctuation, e.g., note:, wasn't parsed properly by DEVONthink Markdown. Fixed.
• Markdown blockquotes separated by one empty line no longer merge into one quote. Fixed.
• The DEVONthink Markdown parser didn't allow CSS declarations containing spaces in YAML frontmatter. Fixed.
• Text in the toolbar search field of the Help viewer could overlap the clear button. Fixed.

• Pro — The MCP server couldn't be accessed during DEVONthink's trial period. Fixed.
• Pro — Looking up reminders via MCP could fail on items having no reminders. Fixed.
• Pro — Items with rich text custom metadata could cause an exception for the MCP server. Fixed.
• Pro — If an invalid value was given while modifying custom metadata via MCP, it could remove data from other fields. Fixed.
• Pro — Downloading a PDF based on its DOI via MCP could fail even if the PDF was publicly accessible. Fixed.
• Pro — Tagging bookmarks with an invalid URL or while being offline could crash when done via Apple's Foundation model on macOS Tahoe. Fixed.
• Pro — On rare occasion, web searching with AI could cause a crash in the Chat assistant, smart chat actions, or the get chat response for message AppleScript command. Fixed.
• Pro — Google's Nano Banana 2 and Pro could fail due to being asked for the wrong type of output. Fixed.
• Pro — Editing an image with Flux Pro or Max could crash if the original image's size was too large. Fixed.

• Server — Long Bonjour names specified in the Server > General settings could prohibit the web server from starting. Fixed.
• Server — PDF documents with emojis or an em-dash in their name downloaded in the web interface had an incorrect filename and a .txt extension. Fixed.

• Sync — Bonjour syncs could cause exceptions on rare occasion. Fixed.

• Pro — To reduce time and token costs of the chat assistant, we removed Allow image generation from the AI > Chat > Assistant settings. AI-generated images can still be made via the Generate Image window.
• Pro — We removed the Style setting from the AI > Image Generation settings and Generate Image window. We also removed the style parameter from the download image for prompt AppleScript command.
• Pro — Google soon discontinues the Imagen models so we removed them from image generation.

Version 4.3 Herschel

DEVONthink 4.3 Herschel introduces our MCP server: A safe, private bridge that lets Claude, Cowork, Codex, and other AI clients work directly with your databases. No scripting or third-party tools required. On macOS Tahoe, we now support Apple Intelligence for light tasks and Apple's faster SpeechAnalyzer for transcription. Other AI improvements include a leaner search assistant, a dedicated tagging model, and web search via ollama.com.

A new Markdown parser and renderer replaces MultiMarkdown 6. It adds callouts, citations, and better compatibility with AI output. Desktop widgets bring the reading list, favorites, and workspaces to macOS Sequoia and later. For DEVONthink Server users, the web interface gains classification, context menu commands for moving, duplicating, and replicating items, and a dialog for opening item links. It also features a much better PDF viewer. As always, reliability and performance continue to improve.

• Replacing MultiMarkdown, our new Markdown parser and renderer supports features like callouts, citations, and tables with captions. It also brings improved CriticMarkup support and is more compatible with AI-generated output. Choose the renderer in the Preview popup in the Files > Markdown settings.
• Supported on macOS Sequoia and later, new desktop widgets let you access your Reading List or Favorites, load a Workspace, or view a tip from our blog.
• View .highland Markdown files used by screenwriters.
• The new URL command, x-devonthink://open-workspace?name=, lets you open an existing workspace.

• Pro — DEVONthink gets its own MCP server that provides a bridge between your databases and supporting AI applications, e.g., Claude. Comprising a suite of almost 60 commands, it is built with data privacy and security in mind, controlling local or remote access, redacting private information from being transmitted, honoring all exclusions from AI set on items (or databases), and limiting access to the internals of your database in the filesystem. In addition, when the server is running, a menu bar icon appears with commands for controlling the server, accessing the settings, or copying the URL you need. Requires macOS Sequoia or later.
• Pro — Use the new Settings > AI > MCP settings to set up and install the DEVONthink MCP server. The options include settings for privacy and making safe connections to other devices, on and off your network.
• Pro — The option Exclude from Chat & MCP in the Database Properties panel lets you easily exclude entire databases from AI access. This is enabled by default for encrypted and revision-proof databases.
• Pro — DEVONthink supports Apple Intelligence's Foundation model in the AI > Chat settings and AppleScript commands on macOS Tahoe and later. At this time, its use is limited to simple tasks like renaming, tagging, etc. as the context window is very small and vision is not supported.
• Pro — Choose the AI model for automatic tagging functions with the new Chat Suggestions popup in the Files > Tags settings. DEVONthink uses this, e.g., when importing files, using the Data > Tags commands, or smart rules and batch processing actions.
• Pro — Added support for LM Studio's reasoning effort (0.4.8 or later).
• Pro — You can now use an ollama.com account for web searches in the AI > Search settings.

• Server — Added an Open Item Link button to the Action toolbar button and the context menu of the item list in the web interface. This opens a dialog to enter an item link to be revealed.
• Server — Added a Reveal command to the context menu of search results in the web interface's item list, revealing the item and its location in the Navigate sidebar.
• Server — Added Move, Duplicate, and Replicate commands to the context menu in the web interface's item list, allowing you to navigate to and choose the destination for these functions.
• Server — Added a section for classifying items to the See Also inspector of the web interface, including a context menu with related commands.

• Converting Markdown documents to plain text, e.g., via the Data > Convert > To Plain Text command, strips all Markdown formatting and converts MathJax and DisplayMath into a more readable form.
• Sheets transcluded in Markdown use an underlying Markdown representation for a more consistent appearance across documents.
• The "What's New" window also appears while DEVONthink is in fullscreen mode.
• Copying item links includes a Markdown-formatted link for use with third-party editors.
• You can use an alternative syntax, :==, synonymously for the is search operator, e.g., in created:==2026-04-14.
• Commands like Open Externally use fewer resources, improving DEVONthink's responsiveness.
• Both the digital object identifier AppleScript property and the Digital Object Identifier smart rule and batch processing placeholder use also the URL to retrieve the DOI. The chat assistant and MCP server of the Pro edition benefit from this too.

• Pro — AI responses use our new Markdown parser, e.g., in the Chat assistant or the summary and transform popovers. Additionally, the Chat assistant knows how and when it's appropriate to use Markdown callouts (admonitions).
• Pro — The Chat assistant, AI smart actions, summarize and transform functions, AI-related AppleScript commands, and the new DEVONthink MCP server, redact or anonymize email addresses, links, credit card numbers, phone numbers, authentication tokens, and labeled secrets in the information sent to LLMs.
• Pro — The Chat assistant displays images returned by Perplexity Sonar models.
• Pro — The Help assistant uses fewer tokens and is faster and more reliable.
• Pro — "Exclude from Chat" in the Generic inspector is now Exclude from Chat & MCP.
• Pro — Local Apple SpeechAnalyzer replaces the "Local Apple Speech" option in the AI > Transcription settings on macOS Tahoe and later. This utilizes Apple's new SpeechAnalysis framework which processes audio locally and up to five times faster. Additionally, it will automatically download and install resources available for the specific language being transcribed.
• Pro — Researching PubMed and arXiv via the Chat assistant is more reliable.
• Pro — If AI requests the text contents of a document, DEVONthink provides a Markdown representation for better access to things like links and styling. This applies to RTF(D), HTML pages, formatted notes, and web archives. (PDF and EPUB documents are already supported). It also improves responses from many AI features, e.g., the Chat assistant and the get chat response for message AppleScript command.
• Pro — Batch tagging is more reliable with low-tier models, e.g., Magistral.
• Pro — The default role for AI-related features, e.g., the AI smart actions, get chat response for message AppleScript command, etc., includes the date, weekday, and current time.
• Pro — If the digital object identifier (DOI) or international standard book number (ISBN) is in the text, metadata, or content of a document, it is available to AI functions like the Chat assistant, Chat - Query smart action, or get chat response for message AppleScript command without processing the document further.
• Pro — Tagging documents with AI retains the spelling when Files > Tags > Use only existing tags of a database is enabled.
• Pro — An improved definition of DEVONthink's search syntax makes database searches via the Chat assistant, Chat - Query smart actions, and the get chat response for message AppleScript command more precise and reliable.
• Pro — Claude Opus is now at version 4.7, both via API and OpenRouter. With a knowledge cutoff of January 2026, it supports up to one million input tokens and 128,000 output tokens as well as high-resolution image processing, up to 3.75 megapixels. The Sonnet models also support up to one million tokens now.
• Pro — Updated OpenAI's GPT models to version 5.5, supporting a context window of one million tokens and a knowledge cutoff date of December 1, 2025, but at a much higher cost. Also, the internal names of these models no longer reference the version, ensuring future updates don't adversely affect scripts, templates, or smart actions.
• Pro — Updated Gemini 3.1 Flash (Preview) to Gemini 3.5 Flash, both via Google and OpenRouter. Note that the new model is three times as expensive as the former one.
• Pro — Gemini 3.1 Flash Lite is available via API and OpenRouter.
• Pro — Updated the Mistral Small model to version 4 and Mistral Medium to 3.5, both with a context window of 256,000 tokens and reasoning capabilities. However, these models are much more expensive than the previous ones.
• Pro — Qwen 3.5, Gemma 4, and versions 3 and 4 of Mistral Small are better supported locally and via OpenRouter.
• Pro — Improved token usage when using Perplexity for Web & Wikipedia searches.
• Pro — Updated the Flux image generation models to Flux 2 Klein, Pro, and Max. They support standard and HD quality and image editing, but not styles.
• Pro — Nvidia's Nemotron 3 is better supported.
• Pro — Xiaomi models are available in OpenRouter.
• Pro — Updated many models in OpenRouter: Grok 4, replaced by Grok 4.2. Mistral Small 3.2, replaced by Mistral Small 4. Gemma 3, replaced by Gemma 4. GLM 4.6 and 5, replaced by GLM 5.2. GLM 4.7 Flash, replaced by GLM 5(V) Turbo. Xiaomi MiMo-V2-Flash/Omni/Pro, replaced with MiMo-V2.5 and MiMo-V2.5 Pro. DeepSeek 3.2 and DeepSeek 3.2 (Speciale), replaced with DeepSeek V4 or DeepSeek V4 Flash. Kimi K2 2.5, replaced by Kimi K2 2.6. All Qwen models, replaced by various 3.6 models.
• Pro — OpenAI's GPT-OSS models are better supported via LM Studio or Ollama local.

• Server — A better PDF viewer in the web interface supports a table of contents, thumbnails, and searching in all browsers. Additionally, toolbar search terms are automatically used when selecting a PDF document in the search results.
• Server — The context menu in the web interface's item list is better supported on mobile devices.
• Server — Double-clicking a group in the search results of the web interface's item list now opens the group.
• Server — DEVONthink's self-signed certificates generated in the Server > General settings use even stronger 2048-bit encryption.

• Sync — Synchronizing RSS feeds less frequently duplicates articles.

• Converting Markdown documents to paginated PDF documents could sometimes fail. Fixed.
• Opening braces with no matching closing braces could break Markdown transclusions. Fixed.
• Converting images to PDF or merging PDF documents on macOS Tahoe could produce a document with only white pages. Fixed.
• Scrolling or selecting pages in a PDF document could mark the document as modified. Fixed.
• On rare occasions, annotating PDF documents could cause an exception. Fixed.
• Multi-page TIFF files couldn't be viewed or converted on macOS Tahoe. Regression. Fixed.
• Web windows opened without a record, e.g., via AppleScript, always opened at the same size with the Inspectors open. Fixed.
• Scrollbars in Widescreen mode didn't behave properly on macOS Tahoe. Fixed.
• Resizing document windows after switching to side-by-side mode didn't work correctly on macOS Sequoia and later. Fixed.
• Group tags didn't appear correctly in inactive tag fields. Fixed.
• In tagging fields, the popup arrow on hovered tags didn't appear in the correct position. Fixed.
• Depending on the date, time, and current locale, the contents of the Log window's Date/Time column could appear clipped. Fixed.
• Tooltip and click positioning in the Editing and Navigation bars were affected by a window's width on macOS Tahoe. Fixed.
• When using a Markdown document in the Annotation field of the Annotations & Reminders inspector, the Format > Add Link and Format > Edit Link commands weren't supported. Fixed.
• The Content > Table of Contents inspector didn't display correctly with Markdown documents using more than three backticks to mark fenced code blocks. Fixed.
• Clearing the color of groups and tags from the Generic Info inspector or Info popover didn't sync between devices. Fixed.
• On startup, a failure to open a database showed an alert that could freeze the application on rare occasions. Fixed.
• The JavaScript command addReminder didn't use the correct scheduling. Fixed.

• Pro — Transcribing PDF documents on demand, e.g., via Data > Recognition or automatically, produced inconsistent results. Fixed.
• Pro — Empty black or white PDF pages weren't ignored as input, leading to unnecessary AI time and token usage on macOS Tahoe. Fixed.
• Pro — Blockquotes in the Chat assistant weren't styled. Fixed.
• Pro — Text pasted into the Title field of the Imprinter settings incorrectly added a % character. Fixed.
• Pro — Adding or deleting a Font placeholder in the Title field of the Imprinter settings didn't update the Font selector popup. Fixed.
• Pro — Inserting a placeholder before an existing placeholder in the Imprinter settings replaced all the text following the new one. Fixed.
• Pro — Custom metadata definitions added via AppleScript could cause sync issues. Fixed.
• Pro — If DEVONthink was not running, the first scan from third-party software launched the app but didn't run OCR on it. Fixed.

• Server — In the web interface, PDF documents couldn't be viewed in Safari. Fixed.
• Server — PDF documents viewed in the web interface on an iPad only displayed the first page. Fixed.
• Server — Markdown documents didn't display images using relative links in the web interface. Fixed.
• Server — The table of contents for EPUB files didn't work. Fixed.
• Server — The arrow indicating the item list's sort direction was the opposite in the web interface. Fixed.
• Server — The user icons in the Server > Users settings didn't display correctly when launching in dark mode and switching to light mode while the settings were open. Fixed.

• Pro — Removed the Remote Apple Speech option from the AI > Transcription settings on macOS Tahoe and later.
• Pro — Removed OpenAI's GPT 4.1 models from API access and OpenRouter. We've also removed GPT 5.1, GPT 5.2, and GPT 5 Nano and Mini from OpenRouter as they have been superseded by newer models.
• Pro — Removed Claude Opus 4.6 (Fast) from OpenRouter.

Version 4.2.2

Today's maintenance release lets you create powerful AI-driven searches and add tags more reliably via scripting. If you are using JavaScript for your automations, you can now use .js files. In AI, Gemini and Nano Banana get updates, and Claude 3 Haiku retires. We also have a few fixes, including better handling of .opml email attachments.

• Pro — Perform AI-driven research with multiple sources, e.g., arXiv, PubMed, even your databases, with the new perform chat research AppleScript command. Control where AI searches with the sources parameter and the number of results with maximum results.

• The appearance of placeholder text in the interface, e.g., in some empty table views, is more consistent with the system settings.
• We made some cosmetic improvements to the sidebar in the Help viewer.
• You can now load external JavaScript for Automation (JXA) scripts with a .js extension in the Scripts menu, as well as use them for smart rules and batch processing.
• Tags separated by commas or semicolons are better parsed and applied when set via scripting.
• Updated application icon for older macOS releases and Sparkle.

• Pro — The automatically generated identifiers of custom metadata (see Data settings) now ignore characters reserved for our search syntax, e.g., :.
• Pro — Database searches via the Chat assistant, Chat - Query smart action, or the new AppleScript perform chat research command now filter irrelevant results.
• Pro — Revisions to the internal names of Anthropic's Claude models ensure future model updates won’t break existing scripts, templates, or smart rules. Also, the AI > Research Report and The Devonian Times templates now use Sonnet's revised name.
• Pro — Gemini Pro and Flash Lite are now at version 3.1 (Preview) via the native API and OpenRouter. Also, Gemini 3.1 Pro and 3.1 Flash handle reasoning better. Note that Gemini Flash Lite 3.1 (Preview) is more expensive than the former version 2.5.
• Pro — More reliable support in the chat assistant, the Chat - Query action of smart rules and batch processing, and the get chat response for message AppleScript command for evaluating mathematical expressions. In addition, the constant eps0 is now also supported.
• Pro — Ollama has improved support for the lfm2 model and GPT-OSS now reasons better.
• Pro — Ollama reasoning models (except GPT-OSS) enable thinking now only when using Settings > AI > Chat: Usage: Best (option is also available in the chat assistant). Otherwise it's disabled to speed up responses (e.g., up to twelve times faster in case of Qwen3.5).
• Pro — Nano Banana 2 replaces version 1 and now supports the Quality (HD or Standard) setting.
• Pro — The script assistant supports GPT 5.3 Codex via OpenRouter.
• Pro — We improved the AppleScript dictionary description of the transcribe record… language command.

• Server — When revisiting the web sharing interface, a warning dialog will appear if you're not logged in.

• Repairing a database with orphaned OPML documents incorrectly converted them to RSS feeds or outlines. Fixed.
• The View > Hide/Show toolbar command wasn't disabled in the Help viewer. Fixed.
• Converting an empty rich text document to plain text or Markdown failed. Fixed.

• Pro — Importing OPML email attachments always converted them to RSS feeds or outlines and could create orphaned files. Fixed.

• Pro — Removed the soon-to-be-retired Claude 3 Haiku. Bulk tasks, e.g., filtering search results, are now done with 4.5 Haiku. Note that the 4.5 model is more expensive.
• Pro — Removed all Qwen3 235b models from OpenRouter, except the ones for coding. Use the latest Qwen 3.5 models instead.

Version 4.2.1

This maintenance release lets you mirror DEVONthink reminders to Apple Reminders via a new script. It also includes a hidden preference to disable dark mode only for PDF documents on macOS Tahoe or later, and lets you run batch processing on specified records via AppleScript. The Pro and Server editions support Claude Sonnet and Opus 4.6 and improve the support for zAI and Qwen models. The update brings several fixes, too, e.g., for lists in rich text documents not displaying on macOS Tahoe.

• The Scripts > Reminders > Export to Reminders menu item mirrors DEVONthink reminders to a DEVONthink list in Apple Reminders.
• A hidden preference, DisablePDFDarkMode, disables dark mode for PDF documents on macOS Tahoe or later.
• The perform batch configuration… record AppleScript command lets you run a batch configuration on specified records.

• Server — In the web interface, the toolbar search now supports the scope: prefix and its abbreviation, s:, similar to the search AppleScript command (see below).

• Search hits in PDF documents are better visible in dark mode on macOS Tahoe.
• Sorting sheets is more reliable on columns with identical cell content, especially dates.
• DEVONthink now better captures web content as clutter-free documents.
• The width of the current main or document window again determines the maximum width of the inspectors pane.
• The Thumbnails inspector is more reliable and should alleviate some rare crashes, e.g., caused after moving documents.
• Updated icon for the Global Inbox counterpart in the Finder.
• The AppleScript search command now supports the scope: prefix (or its abbreviation, s:) if you leave the in parameter unspecified or as a missing value. Supported scopes are: all, inboxes, inbox, tags, trash, or the quoted name of a database.

• Pro — Updated support for Claude Sonnet and Opus to version 4.6 via the native API and OpenRouter. The script assistant also uses the latest version of Claude Sonnet. For both models, DEVONthink uses adaptive thinking and effort instead of budget tokens (only natively via Anthropic).
• Pro — The AI > Research Report - PDF, AI > Research Report - HTML, and AI > The Devonian Times templates now use Claude Sonnet 4.6.
• Pro — Replaced zAI's GLM 4.7 with the latest version GLM 5 from OpenRouter.
• Pro — Improved support for zAI's GLM 4.x models.
• Pro — Optional support for API keys for LM Studio 0.4.0 or newer.
• Pro — Updated or replaced various OpenRouter models for the script assistant with more powerful models. Additionally, we have removed several OpenRouter models that no longer exist.

• Lists, e.g., bulleted lists, in rich text documents didn't display correctly on macOS Tahoe. Fixed.
• Checkboxes in sheets, as well as boolean custom metadata in List views in the higher editions, didn't display correctly on the latest macOS releases. Fixed.
• In the Info > Annotations & Reminders inspector, text colors in Markdown annotation documents were inconsistent. Fixed.
• Dates in sheets or List view columns weren't parsed correctly if the date contained a four-digit year and didn't match the format set in the system settings. Fixed.
• PDF text and note annotations didn't always remember their last used font, so changes such as moving or editing text reverted them to a default font. Fixed.
• Clipping paginated PDF documents didn't build them correctly on macOS Tahoe. Fixed.
• Emptying the trash could show a wrong, unnecessary alert in case of indexed, trashed items which can't be deleted (e.g., located in the user's library folder). Fixed.
• Finder tags of updated indexed items were used if the option to export Finder tags (see Settings > Files > Tags) was enabled instead of the option to import Finder tags. Fixed.
• Removing all tags, e.g., via Data > Tags > Remove All Tags, didn't trigger the On Tagging event of smart rules. Fixed.
• Minor visual glitches of File > Database Properties popover in the case of read-only databases. Fixed.
• The createRecordWith JXA command didn't support all properties of smart groups. Fixed.
• The unlinked Wiki links AppleScript property was incorrectly included in the properties property. Fixed.
• The direct, unnamed parameter of the search AppleScript command wasn't marked as required and can't be a missing value. Fixed.

• Pro — Settings > Imprinter didn't always save modified options. Fixed.
• Pro — Changing the duplex setting of the scanner interface could freeze the application. Fixed.

• Pro — Removed unsupported GPT Audio and GPT Audio Mini models from OpenRouter.
• Pro — Removed obsolete MoonshotAI Kimi K2 0905 and MoonshotAI Kimi K2 Thinking models from OpenRouter. Use the latest MoonshotAI Kimi K2.5 model instead.
• Pro — Removed obsolete zAI's GLM 4.5 Air model from OpenRouter. Use the new GLM 4.7 Flash model instead.
• Pro — Removed Qwen3 Max and Qwen3 VL models from OpenRouter. Use the latest Qwen3.5 models instead.

Version 4.2 Cassini

DEVONthink introduces a refreshed look with a new app icon and full support for Apple's Liquid Glass design on macOS Tahoe. On macOS Sequoia and later, sidebars and inspectors now animate smoothly in and out, and the Navigate sidebar is spring-loaded. Simply drag items to the window edge to reveal it automatically. PDF documents on Tahoe also respect your dark mode preference for more comfortable reading.

The Chat assistant gains powerful new capabilities: adjust many options directly within the chat window, and access up to 50 saved conversations to pick up right where you left off. You can now choose between DEVONthink, Perplexity, or Exa for web and Wikipedia searches, giving you control over speed versus freshness. Updated AI models including Claude 4.5, GPT 5.2, Gemini 3, and Google Nano Banana Pro, deliver improved performance across all AI use. When generating images, you can choose the aspect ratio.

New workflow tools help you work faster: the detachable Graph popover lets you visualize document connections, you can manually save named versions for finer revision control, and annotation files are now easier to create, locate, and open directly from menus and inspectors. Markdown documents also support smart quotes and dashes for polished typography.

• DEVONthink supports Apple's Liquid Glass design language on macOS Tahoe and has a new, fitting app icon. We have modified many parts of the user interface and will continue to do so while Apple addresses issues. Many refinements we've made, e.g., to the popovers, apply to all supported system versions.
• The Data > New > Sheet panel lets you name the new sheet.
• An Open button in the Annotations & Reminders > Annotations inspector lets you quickly open the current annotation file.
• The Versions inspector supports manually creating named versions.
• The settings option Files > Markdown: Preview: Smart quotes and dashes tells DEVONthink to render quotes and dashes using their typographically correct variants.
• A warning shows when a database may be open in another user account.
• After updating DEVONthink to the next feature release, a What's New panel shows the top features of the release.
• A hidden preference, WindowToolbarTitleHidden, lets you hide a main window's title and gives you more space for toolbar items. Note that you cannot use it with the WindowToolbarExpanded hidden preference.
• The unlinked Wiki links AppleScript property for records helps detect WikiLinks without a linked item.
• The HTTP response class has two new AppleScript properties, error code and error domain, to aid in troubleshooting.

• Pro — New settings in the Chat assistant let you dynamically change options for the current chat, including the Search scope, Usage, and Documents mode.
• Pro — The Chat assistant now saves up to 50 chats, defined via the option Max. Recent Chats in the AI > Chat settings. DEVONthink automatically names previous chats and lets you access or remove them via the button in both the Chat inspector and popover. Up to ten of today's chats are listed at the top with previous chats in date-specific submenus. Selecting a previous chat will attempt to reveal or open the associated document or selection with the previously used document mode.
• Pro — You can open the Chat popover via the keyboard shortcut ⌃︎⌘︎A that we added to Tools > Chat.
• Pro — The new menu command Tools > Graph opens the Graph inspector in a floating, resizable popover. This popover will be restored on launch and with workspaces.
• Pro — Separate AI > Search settings let you control where you want AI to search. This also includes new Web > Wikipedia settings to search via DEVONthink (slower but more up to date), Perplexity, or Exa (faster, but using indexed content).
• Pro — The AI > Research Report template is now also available in a PDF variant.
• Pro — Support for DeepSeek 3.2 and 3.2 (Speciale) via OpenRouter.
• Pro — Support for the image generation model GPT-Image-1-Mini, accessible in the AI > Image Generation settings and the Generate Image panel.
• Pro — Support for Google Nano Banana Pro, generating images in 2K and 4K. Note that it generates images more slowly and is more expensive than Google Nano Banana. Also, the Mixture-of-Agents - Images template uses Google Nano Banana Pro, if available.

• Main and document windows now use modern APIs on macOS Sequoia and later, providing animated toggling of the sidebars and inspectors. You can also collapse these panes by dragging the separator to the window's edge.
• On macOS Sequoia and later, the Navigate sidebar is now spring-loaded and will automatically pop open if you drag an item to the left side of a main window.
• The Format Bar is available in fullscreen window mode on macOS 15 and later. You show or hide it with Format > Show/Hide Format Bar.
• The selected row in the Move To and Insert Link popovers has better contrast when they don't have focus.
• DEVONthink more consistently positions popovers relative to the current window.
• You can make the group selector larger than before.
• The appearance of informational text is more consistent throughout the interface.
• On macOS Tahoe, PDF documents now display in dark mode when General > Appearance > Use dark background for documents is enabled.
• When converting a rich text document to PDF, DEVONthink checks if the source document only contains a single PDF attachment and no useful text, then uses the found PDF. This also applies to email conversions in the Pro and Server editions.
• New entries made via Data > Add To > Table of Contents can use the bounds of a selection instead of just the page.
• The Table of Contents inspector is more compatible with poorly constructed HTML content, e.g., created by some AI models.
• The "Remind Me" submenu in the Data and context menus is now Annotate & Remind. You can still set a due date on a selected item but you can also create a new annotation document via the template of your choice. The annotation will open in a new document window or be displayed in the Annotations & Reminders > Annotations inspector, if it's open.
• The Insert Quote command of the Info > Annotations & Reminders inspector supports also Markdown documents.
• In the AI > See Also inspector, the keyboard shortcut for the Move To command is only shown when the ⌃︎ Control key is pressed.
• The context menu of the item list of main windows now includes a Reveal command to reveal replicants, duplicates, and an item's annotation document.
• The command Apply Rule is now available in the context menu for smart rules added to the Favorites.
• The document date placeholders submenus are now under the Document options.
• Using Data > Mark to lock or unlock a group applies the same to its children. This does not apply when locking an item via scripting or smart rules.
• Drag-and-drop or copy-and-paste is now more compatible with third-party editors.
• DEVONthink no longer shows a warning dialog after it crashed or was quit forcefully.
• Automation prompts to allow DEVONthink to control the Finder no longer self-dismiss.
• The compare AppleScript command now only returns a maximum of 40 results, consistent with the See Also & Classify inspector.
• You can now control how long the download URL and download markup/JSON from AppleScript commands run by using a with timeout of x seconds… end timeout block.

• Pro — Bookmarks are now supported in many AI-driven functions, like labeling, rating, and tagging via the Data menu, summarization, as well as smart rules, batch processing, and AI placeholders.
• Pro — When using an AI model supporting tool calls, e.g., the ChatGPT models, the Chat assistant can temporarily highlight words or sentences in the currently viewed document, similar to the Search inspector.
• Pro — The Chat assistant can now create comma or semicolon-delimited sheets (.csv), instead of just tab-delimited (.tsv) ones.
• Pro — The "Research Overview" template is now AI > Research Report - HTML.
• Pro — The AI > Research Report and AI > The Devonian Times templates use updated models: Gemini 3 Pro or Flash Preview, GPT 5.2, or GPT 5 Mini.
• Pro — The Help assistant now first tries to determine the most relevant help pages. This can drastically reduce the number of tokens used, up to 90%, and allows use of newer or more expensive (and more capable) models.
• Updated Google Nano Banana from preview to final release.
• Pro — Aspect Ratio replaces the "Size" option in the AI > Image Generation settings and Generate Image panel. Also, the aspect ratio AppleScript parameter replaces the "size" parameter of the download image for prompt command. Supported sizes are 1:1, 16:9 (9:16), and 4:3 (3:4). This setting as well as which generative model is used and, optionally, the quality setting, control the size of the resulting image.
• Pro — The Chat assistant and AI actions like tagging or rating, better handle errors. The assistant also more reliably clears chats. DEVONthink better filters AI-generated error messages like those from placeholders in smart rules and batch processing, and also handles OpenRouter errors better, e.g., when a provider is overloaded.
• Pro — Updated GPT 5.1 to GPT 5.2, natively and via OpenRouter. Also, GPT 5 Pro, 5.1 Codex, and 5.1 Codex Mini are available in the Chat assistant via OpenRouter, the Codex and Codex Mini models in the Script assistant.
• Pro — Updated GPT-Image-1 to GPT-Image-1.5. The Data > New From Template > AI > Mixture-of-Agents template support the new model.
• Pro — Updated Anthropic Claude 4.1 Opus to 4.5 and Haiku to version 4.5, natively and via OpenRouter. Claude 4.5 Opus is less expensive, a new parameter reduces response time and costs. Claude 4.5 Haiku supports reasoning, more output tokens, and has knowledge up to July 2025.
• Pro — Updated Gemini 2.5 Pro to Gemini 3 Pro and Gemini 2 Flash to Gemini 3 Flash (Preview), natively and via OpenRouter. Also, responses to vision requests are quicker, cheaper, and support a vision limit of 200 items, useful for video, multi-page PDF documents, and multiple selected images.
• Pro — Updated support for Mistral Large 3, including via OpenRouter. Also Mistral's Devstral Medium model is now Devstral 2, with a context window of 256,000 tokens and supported in the Script assistant.
• Pro — Mistral and local models have improved tool call support.
• Pro — Ollama, LM Studio, and OpenAI (Compatible) models now require a minimum context window of 4,096 tokens. We recommend 16k (16,384), when possible.
• Pro — Depending on the usage setting, LM Studio can process images with a maximum dimension up to 2,048 pixels.
• Pro — DEVONthink examines and filters OpenAI (Compatible) models for capabilities, e.g., vision or tool call support, with servers not reporting this information on their own. Note that the context window still needs to be specified. This provider is also more compatible with endpoints hosting models like GPT 5.x or the O-models.
• Pro — Updated further models usable via OpenRouter: Grok 4 Fast to Grok 4.1 Fast, Nova Lite 1.0 to Nova 2 Lite, and zAI's GLM to version 4.7.
• Pro — Better support for the Qwen3 models.
• Pro — OpenRouter models now prioritize price (Cheapest) or thoroughness (Best) for models having multiple providers, depending on the AI > Chat > Usage setting.
• Pro — The Chat - Query smart action works more reliably with low-tier models like GPT 5 Nano or Mini.
• Pro — The get chat response for message AppleScript command has a better description.
• Pro — The as parameter for the get chat response for message AppleScript command accepts PDF to return a paginated PDF document of the response.
• Pro — The appearance of the OCR settings is more consistent with other settings panes.

• Sync — Bonjour passwords stored in Keychain Access now use the host's ID as the host name of a device can sometimes change when switching networks.

• Ordered lists in Markdown didn't render properly if there was no space after the dot. Fixed.
• Inline MathJax equations didn't always render properly in Asian language documents. Fixed.
• Web pages using the Consent framework crashed WebKit. Fixed.
• Converting emails (Pro and Server editions) and rich text documents to PDF documents could stall the app. Fixed.
• Converting HTML documents didn't convert code blocks correctly. Fixed.
• Capturing a web page within DEVONthink didn't honor the clutter-free setting. Fixed.
• Searching for tags in a toolbar search or via smart groups and rules didn't work correctly with groups excluded from tagging and grouped RSS feeds. Fixed.
• Switching between light and dark modes could cause visual artifacts in labels, ratings, and the Activity pane. Fixed.
• Destination popups, e.g., in the Sorter, didn't display correctly in dark mode if the destination was the root of the Global Inbox or database. Fixed.
• When detached, the Information popover didn't update when switching tabs in the frontmost document window. Fixed.
• The header of tables in various inspectors, e.g., Concordance > List showed artifacts on macOS Tahoe. Fixed.
• For favorited groups, the item count wasn't shown and their unread count didn't match what was reported in the item list. Fixed.
• The internal versions group was neither properly filtered when determining statistics in Database Properties nor was it ignored in toolbar searches. Fixed.
• Deleting a version from the Versions inspector would incorrectly close the current tab or document window. Fixed.
• The Tools > Show Info command closed the popover instead of activating it. Fixed.
• The context menu could display an unnecessary and disabled Put Back command. Fixed.
• The keystroke ⌃︎↩︎ on modern versions of macOS didn't work properly in Icon view and could also change text selections. Fixed.
• Pressing the ⎋︎ Escape key with the database authentication panel open caused an issue. Fixed.
• The Sorter didn't always appear in the menubar on macOS Tahoe. Fixed.
• The Apply Rule smart action didn't work properly if the applied rule ran asynchronous actions, like chat suggestions. Fixed.

• Pro — In rare cases, while in List view with a specific custom metadata field shown, disabling that field on another device and synchronizing could crash the app. Fixed.
• Pro — The Tools > Chat command didn't close an already open popover. Fixed.
• Pro — The prompt field didn't have focus when the Chat popover opened. Fixed.
• Pro — The Model popup in the Script assistant was too narrow for many OpenRouter models. Fixed.
• Pro — If the same model was accessible via a local and a remote Ollama installation, the local model was chosen in the Model popup in the Chat assistant or AI-driven smart actions, e.g., Chat - Query. Fixed.
• Pro — Embedded models in Ollama were not filtered from the model list. Fixed.
• Pro — The Chat toolbar item wasn't disabled when a Chat popover was already open. Fixed.
• Pro — Markdown blocks in chat messages were sometimes stripped out during chat or when saving the conversation. Fixed.
• Pro — A document containing text provided to the smart actions Chat - Query and Chat - Continue if wasn't handled correctly if the document mode was set to Without Document or Document (Vision). Fixed.
• Pro — The Data > New > Script window had no title. Fixed.
• Pro — The Quality setting in the Data > New > Generate Image panel wasn't used. Fixed.
• Pro — Using Data > Move to Trash could have caused glitches if the icon view of the scanner interface was active. Fixed.
• Pro — Tags were not applied to a document when scanning to a binder. Fixed.
• Pro — The AI commands in the Rating, Label, and Tags submenus of the Data menu incorrectly showed a star when running the Pro or Server editions. Fixed.
• Pro — Changing the type of custom metadata fields, e.g., from Boolean or a number to a string, then editing a smart rule using these fields, could crash the app. Fixed.

• Pro — Removed tool calls for various AI functions from the Chat assistant: for compiling scripts, support prompts, and for showing a toolbar search. All these functions were superseded by the Search, Help, and Script assistants. These are less expensive, have shorter response times, and are more reliable.
• Pro — Removed Claude Sonnet 4 and 3.5 Haiku, both replaced by the 4.5 models.
• Pro — Removed the deprecated Sonar Reasoning model.
• Pro — Removed the outdated Dall-E 3 image generation model.
• Pro — Removed Grok 4 Fast from OpenRouter, replaced by Grok 4.1 Fast.
• Pro — Removed Devstral Small and Medium, including from OpenRouter, replaced by Devstral 2.
• Pro — Removed DeepSeek 3.1 Terminus from OpenRouter.
• Pro — Removed the Exacto models from OpenRouter as they weren't as reliable as expected, as well as support for Voxtral, GPT-5-image-1, GPT-5-image-mini-1, o4-mini-deep-research, and o3-deep-research.

• Visual glitches on macOS Tahoe after switching between light and dark mode, a common Liquid Glass issue.

Version 4.1.1

This release helps mitigate AI risks, especially when processing online content, to reduce the potential for malicious prompt behavior. We also updated the supported AI providers, adding support for Google's Nano Banana image engine and Anthropic's Claude 4.5 Sonnet.

DEVONthink now detects empty pages when clipping PDFs, ensuring more consistent and useful results, and we made the Safari and share extensions more reliable on macOS Tahoe. Mailboxes in the Import sidebar load faster and match Apple Mail's appearance and you can modify a document's searchable text via the chat or scripting. On the sync side, DEVONthink now automatically adjusts resource usage based on your machine's temperature and workload, and sync in general is more performant. Of course, we fixed bugs and glitches, too.

• We added independent buttons for the Insert Back Link and Insert Quote commands to the Annotations & Reminders > Annotations inspector.

• Pro — We added two new options to the AI > Chat settings: Allow download of web pages and Load remote images automatically to let DEVONthink download and examine web content while generating a response. We disabled these options by default, as they could be considered insecure.
• Pro — We added separate Destination settings for image recognition and multimedia transcription.
• Pro — DEVONthink now suports Anthropic's Claude 4.5 Sonnet model with pricing and possibilities identical to Claude 4 Sonnet. In addition, the Script Assistant uses now this model instead of Claude 4 Sonnet.
• Pro — We added support for Google's Nano Banana text-to-image engine in all image creation functions, e.g., the Generate Image window. This model supports image editing, and, unusually, defines the size and style directly from the prompt. You must have a Gemini API key to access this model.

• We improved the consistency of syntax colorizing for quotes in Markdown documents.
• When converting EPUB documents, clipping web content, or using the download PDF document from AppleScript command to create PDF documents, DEVONthink now detects empty black and white pages and either filters them out or aborts the clipping.
• When you select a local or global smart group or smart rule in the Navigate sidebar, DEVONthink now uses the targeted group as the destination for new items and grouping selected items.
• We reduced resources used for versioning by 90%, especially during toolbar searches and while updating smart groups and rules.
• DEVONthink now explicitly registers its extensions to ensure they're available in Safari and macOS' Share menu on macOS Tahoe or later.
• You can now modify the plain text AppleScript property of transcribed files, e.g., images, PDF documents, audio and video files. You can also use the transcription synonym.
• The external smart script Zettel-ID handles errors caused by Markdown documents that don't contain this ID.
• We improved the handling of menu scripts execution.

• Pro — Apple Mail mailboxes and their structures now load faster and more reliably in the Import > Emails sidebar.
• Pro — Cross-linking items is now easier; just drag and drop an item onto the item link pop-up menu in the Info > Data inspector or sheet form views.
• Pro — The Help assistant provides the current help page to the chat, enabling summarization, translation, and other AI-assisted tasks.
• Pro — The Help and Chat assistants retain unsent prompts when reopening windows, inspectors, or popovers.
• Pro — A revised system prompt for chats eliminates unnecessary AI remarks about anonymized links.
• Pro — You can optionally save images created or edited via the Chat assistant to the database via prompt. DEVONthink adds the prompt to the Finder Comments of the image.
• Pro — The Chat assistant lets you modify the searchable text of transcribed documents (e.g., images, PDFs, audio, and video files), allowing updates to translated or spell-checked versions.
• Pro — The context window of the current AI model now determines the number of tags available for the Files > Tags > Import > Use only already existing tags of a database setting.
• Pro — Database searches are faster across the board, from the Chat assistant to smart actions and scripting, with notable gains for local AI models.
• Pro — Web and Wikipedia searches via the Chat assistant, AI smart actions, and AppleScript commands now also use the Ecosia and Yahoo search engines and better filter results from Bing and Brave.
• Pro — All AI responses, whether from summarization, transformation, script commands, or the Chat assistant, now undergo sanitization to remove potentially insecure content. Tool call names are stripped from the input, web page downloads are restricted to non-document queries, and the Chat assistant disables downloads after accessing item contents, annotations, or comments.
• Pro — DEVONthink more reliably handles broken links and item links, e.g, returned by OpenAI's gpt-oss model.
• Pro — Google's Gemini models now automatically use the latest versions of their Flash Lite, Flash, and Pro models
• Pro — The The Devonian Times and Research Overview templates now default to Claude 4.5 Sonnet when Claude is the selected AI provider.
• Pro — All Perplexity Sonar models now support vision capabilities.
• Pro — The Magistral Small and Medium models use version 1.2 now, adding vision support and expanding the context window to 128,000 tokens.
• Pro — Tool calls with Mistral and local AI are simplified and improved.
• Pro — Ollama (Remote) now supports API keys, e.g., for the new Ollama Cloud.
• Pro — The OpenAI (Compatible) option now works with a broader range of providers and models, including those managed by universities and other institutions.
• Pro — Qwen Code Plus is now available via OpenRouter in the Script Assistant.
• Pro — The download image with prompt AppleScript command clearly reports an error if the provided image data is not an image.
• Pro — We improved the as property descriptions for the chat response for message and add custom meta data commands.

• Sync — The Reading List sidebar and custom metadata now sync more reliably via Bonjour.
• Sync — DEVONthink now detects the battery and thermal state of the machine and automatically throttles CPU and I/O load during synchronization when necessary.

• Thumbnailing many small documents could sometimes cause a crash. Fixed.
• Converting rich text documents or emails to paginated PDF documents could intermittently freeze on macOS Tahoe. Fixed.
• Invalid relative image and transclusion links in Markdown could break document previewing or conversion, or crash the application. Fixed.
• Links entered in lists in Markdown documents didn't work properly in WYSIWYG mode. Fixed.
• Markdown headings containing code didn't display with correct colors. Fixed.
• Links pasted into Markdown lists did not use the correct color. Fixed.
• Searches with proximity operators didn't always highlight hits in PDF documents. Fixed.
• In macOS dark mode, text entered in new HTML documents didn't invert to white. Fixed.
• Tools > Summarize Annotations and the summarize annotations of records AppleScript command sometimes returned annotations in the wrong order, such as those made by PSPDFKit. Fixed.
• Web clipping via the Sorter when opened with the keyboard shortcut incorrectly included Safari's profile name. Fixed.
• Images dropped or pasted into Markdown documents with the Files > Markdown > Import images to group setting enabled didn't apply the Files > Import > Recognition or Files > Tags > Import settings. Fixed.
• Documents moved or duplicated between databases incorrectly added the extension to the internal name and document could be renamed with an empty name if General > Appearance > Show filename extensions was enabled. Fixed.
• Duplicate notifications from the Extras sidebar occurred when multiple main windows were open. Fixed.
• The AppleScript dictionary incorrectly defined and described the root of main windows. Fixed.
• The internal name of files created via AppleScript incorrectly retained the extension. Fixed.

• Pro — Loading the Model pop-up menu in the AI > Chat settings could sometimes cause a crash with local AI providers. Fixed.
• Pro — Claude AI wasn't working anymore in the Help assistant. Fixed.
• Pro — Retrying a chat in the Help assistant failed after reopening the panel. Fixed.
• Pro — The Chat assistant and chat response popovers, such as Summarize, didn't render code blocks or multiple tildes correctly. Fixed.
• Pro — The Reveal command in the item link pop-up menus of the Data inspector or sheet form view always opened a new window. Fixed.
• Pro — AI responses in the Chat assistant, AI smart actions, or the get chat response for message AppleScript command could stall if required items became unavailable, such as when an item was deleted or its database closed. Fixed.
• Pro — Image engines supporting image editing stopped working after creating two images. Fixed.
• Pro — The OpenAI (Compatible) option in the AI > Chat > Provider settings incorrectly required /v1/ in the URL. Fixed.
• Pro — In the German localization, the OCR > PDF Resolution setting couldn't be changed. Fixed.
• Pro — In batch processing, the popovers in the Chat - Query and Chat - Continue if smart actions failed to initialize correctly after switching configurations. Fixed.

• Sync — Bonjour synchronization could sometimes cause memory issues. Fixed.

• Pro — We replaced no longer functioning models and outdated models from the OpenRouter options.

Version 4.1 Europa

Here it is, the first feature release for DEVONthink 4. Choose from a much wider range of server-based AI models via OpenRouter or by connecting to just any OpenAI-compatible provider. Integrate models from, e.g., Amazon, Meta, Microsoft, xAI, Baidu, Moonshot AI, Qwen, and more. In addition, DEVONthink Europa adds native support for Claude 4.1 Opus, Gemini 2.5 Flash Lite, OpenAI GPT-5, Google Imagen 4, and the latest models by Mistral. Connect to your institution's Ollama installation with a new configuration option.

To further automate your work, describe your task to the AI-powered script assistant. It generates AppleScript and JavaScript for Automation (JXA) scripts and even corrects common scripting issues. Edit images, both generated or dropped in from other sources, with prompts using common AI image generation models. As generative AI is notoriously bad at maths, the Chat and other AI-assisted functions now support mathematical functions, and if you're working in academia, retrieval-augmented generation (RAG) gets direct access to the huge document collection on arXiv. In addition, DEVONthink Europa comes with a new Quick Look extension for previewing Spotlight results on macOS Sequoia and later, and more responsive editing of especially larger Markdown documents. It also updates and sorts lists and empties the trash significantly faster.

• A new script editor, opened via Data > New > Script lets you draft different DEVONthink scripts, e.g., for reminders or smart rules.
• A new Quick Look plugin lets you preview documents in, e.g., Spotlight searches on macOS Sequoia and later. This only supports standard file formats: rich and plain text, PDF, HTML, XML, images, audio, and video.
• A new search criterion Kind is PDF+Text serves as a shortened form of Kind is PDF/PS and Word Count > 0. Additionally, the toolbar search syntax is kind:pdf+text, its negated form kind:!pdf+text.
• Added a placeholder to the alarm text field in the Info > Annotations & Reminders inspector.
• The Zettel-ID external script for the Script with Input/Output smart rule action pull this information from Markdown metadata.
• The AppleScript record class supports the new read-only properties language and markdown source.

• Pro — Support for OpenRouter is available in the AI > Chat settings and the AppleScript suite. This provides access to some AI models that aren't natively supported. Currently accessible models are from: Amazon, Anthropic, Google, Meta, Microsoft, Mistral, OpenAI, Perplexity, xAI, Baidu, Deepseek, Moonshot AI, Qwen, zAI, noting we aren't providing access to beta, preview, experimental, outdated, or distilled models. The Help assistant, as well as the Research Overview and Mixture of Agents - Text/Markdown templates, also support OpenRouter.
• Pro — Added support for OpenAI's GPT-5 models. Currently much less expensive than the 4.1 models, it does have a much smaller context window of 400,000 tokens but supports more output tokens and has been trained on more up to date information. Due to the lower costs, web searches are filtered with GTP-5 Nano.
• Pro — Added support for Anthropic's Claude 4.1 Opus.
• Pro — Added support for Mistral's Voxtral transcription to the AI > Transcription settings. Currently, this uses Voxtral Mini and is less expensive than OpenAI's speech-to-text models.
• Pro — Added support for Google's Imagen 4 Ultra in the AI > Image Generation settings, the Generate Image window, and accessible via the download image for prompt AppleScript command. A Replicate.com account is required to use this model.
• Pro — Added vision support to the Perplexity models Sonar, Sonar Pro, and Sonar Reasoning.
• Pro — Added Ollama (Remote) the Chat > AI > Provider settings and renamed "Ollama" to Ollama (Local). This allows you to access Ollama running on a network, e.g., at a company or university. The remote provider is also supported in AppleScript, e.g., by the get chat response for message command.
• Pro — Added a script assistant, available in the Script Editor window, smart actions Apply Script and Script with Input/Output, or reminder scripts. Describe the action you need and a code-compatible AI model will provide a draft script to get things started.
• Pro — Save all images in a chat with the Save All Images context menu of the Chat assistant, available if the current chat contains more than one image.
• Pro — The Generate Image window can now modify a generated image or one dropped or pasted into the window. Also, images generated via the Chat assistant can be modified. And as the Chat popover is persistent between launches, you can modify an image in it at a later time. Currently, this function is only available with certain models: Flux Schnell, Pro, Pro Ultra, Stable Diffusion Large or Turbo, and GPT-Image-1.
• Pro — The Generate Image window now displays a history of created images.
• Pro — Given the poor calculations produced by LLMs, a new tool lets the Chat assistamt, smart actions, and AppleScript commands evaluate mathematical expressions.
• Pro — Added the new The Devonian Times template, summarizing today's news via AI.
• Pro — Added the arXiv site as a search option in the AI > Chat settings.
• Pro — Added a Language menu to the AI > Summarization settings.
• Pro — Added a new built-in smart rule, Apply Chat Suggestions. This applies seven common AI suggestions to matched documents, e.g., tagging, summarization, etc., utilizing placeholders and one AI request. This rule can be duplicated and modified to suit your needs.
• Pro — The Chat assistant, smart actions like Chat - Query, and the download chat response for message AppleScript command can download links in various formats (HTML, Markdown, text, snapshot). The format is automatically chosen, either by the AI or depending on the user's prompt. Currently, this only works with commercial models.
• Pro — The new AppleScript property current chat model returns the default model.
• Pro — The new get chat capabilities for engine… model AppleScript command retrieves the capabilities of a specific AI model.
• Pro — The get chat response for message AppleScript command now has a usage parameter to improve results or control the costs of using external AI.

• "Audit-proof" databases have been renamed to Revision-proof.
• Editing very long Markdown documents is more performant. DEVONthink detects links in Markdown documents at least 25 times faster when syntax coloring and WYSIWYG is enabled.
• Markdown WikiLinking now supports [[link|-]] to use the title of a linked item. This can be useful when using short aliases for linking.
• Prism.js renders Objective-C code marked with the alternate objc or language-objc as well as the default language-objectivec.
• DEVONthink indexes web-based documents and converts them to plain text better.
• The Go > To Document and Edit > Insert > Item Link popovers now search after a short delay to avoid interfering with typing.
• Updating or sorting by name or kind in item lists is up to ten times faster, including in popovers such as Go > To Document.
• Editing annotations in the Info > Annotations & Reminders inspector now supports versioning.
• The Data > Move > Put Back command as well as its context menu counterpart display the destination when only one item is selected.
• For quicker sorting, the Ascending and Descending commands are now at the top of the View > Sort menu and the sort popup menu in the Information bar of main windows.
• The Add and Preview buttons in the Sorter now display their shortcut keys when you hold the Command key.
• In the Sorter's Voice Note and Video Note tabs, the camera and microphone menu is disabled if DEVONthink is disabled in macOS' Privacy & Security > Camera settings.
• Emptying the Trash is up to five times faster when versioning is not enabled.
• DEVONthink's built-in folder action scripts now run smart rules using the perform smart rule AppleScript command with the On Import and On OCR event triggers, as appropriate.
• Querying tags via AppleScript is now supported, e.g., tags contains "support" or where its tags contains {"2025","documentation"}.
• Records now support the sheet-related AppleScript commands get cell at row/column, set cell at row/column, add row cells, and delete row at position.
• The AppleScript database property "audit proof" is now revision proof. However, audit proof has been added as a synonym to ensure backwards-compatibility.

• Pro — Summarizing documents to annotations, e.g., via Edit > Summarize via Chat or Tools > Summarize Documents via Chat, first saves a version when versioning is enabled.
• Pro — The Research Overview template now supports local AI engines and models having tool call support and a context window of at least 32,000 tokens.
• Pro — The notification of a chat-summarized RSS article is simpler when only one article is involved.
• Pro — The search assistant now accepts local AI models with a context window of at least 8,000 tokens.
• Pro — In the Import > Image Capture sidebar, the Group menu now has a search field for quickly locating groups. Also, the destination settings are saved after each scan and restored when reconnecting to the scanner.
• Pro — The From date in the Import > Emails sidebar now defaults to the last used To date.
• Pro — The Chat assistant better handles detecting selected documents, especially when using the Chat popover.
• Pro — AI-directed features using web, Wikipedia, and PubMed searches are more efficient, filtering out excess or links already processed by concurrent searches. This affects e.g., the Chat assistant, smart actions, and AppleScript commands, and could speed up responses and minimize token usage.
• Pro — The Chat assistant's progress indicator more clearly shows chat activity, e.g., searching or downloading resources.
• Pro — The Chat assistant doesn't scroll a received chat response out of sight anymore.
• Pro — Prompt fields across the application, e.g., in the Chat assistant, the Generate Image window, and the Help assistant, now support using ⇧︎↩︎ in addition to ⌥︎↩︎ for inserting a line break.
• Pro — The model menu in the Chat assistant features a search field if more than 25 items are available.
• Pro — Search request made to Google, Bing, and Brave via chat functions, e.g., the Chat - Query action or prompts in the Chat assistant are more reliable. DEVONthink also handles Mistral's rate limits better.
• Pro — Bulk processing items via chat, e.g., tagging or summarizing, uses better error handling.
• Pro — Chat tools calls for local AI models have a reduced temperature for more reliable results and can now download web pages as needed. In addition, the Chat assistant handles errors better.
• Pro — Irrelevant PubMed and arXiv search results are now filtered in all AI functions, e.g., in the Chat - Query smart action, maintaining the context of the current conversation. For commercial providers, the least expensive model is used for filtering. Additionally, once enough results have been received from these site and Wikipedia, other pending evaluations are cancelled, saving both tokens and time.
• Pro — Chat suggestions in smart actions, e.g., suggested tags, are more reliably retrieved.
• Pro — The download image for prompt AppleScript command now reports an error if the specified model doesn't support image editing. When using the Flux engine, the Flux Kontext models are used as they are built for image editing, not just generation.
• Pro — Updated Gemini 2.5 Flash to the final version.
• Pro — Both filtering web search results and the Help assistant use now Gemini 2.5 Flash Lite.
• Pro — Updated Google Image 4 to final version and is used in the Mixture-of-Agents - Images template.
• Pro — Both AI > Mixture-of-Agents templates no longer use Perplexity due to an increase in irrelevant responses.
• Pro — Updated Mistral Medium to version 3.1. Mistral Large 3 has better vision support. And "Mistral AI" has been renamed to Mistral in the AI settings and Chat assistant.
• Pro — The default context window set by DEVONthink for Ollama is now 16,000 tokens.
• Pro — The "GPT4All" provider has been renamed to OpenAI (Compatible) and can be used connect to any compatible remote services. It also now supports an API key, affecting the AI settings, menus and related AppleScript commands.
• Pro — Improved compatibility of chat-based smart actions with LM Studio's MLX models, e.g., Gemman 3n.

• Server — Improved HTTP handling, should also address a rare crash.

• Markdown lists didn't render properly when indented with one space. Fixed.
• Thumbnailing Markdown documents could cause a rare crash. Fixed.
• Scrolling Markdown documents was affected after pasting copied text, changing margins, or syntax highlighting. Fixed.
• Remote images in Markdown documents didn't render in WYSIWYG mode if the files didn't have a valid file extension. Fixed.
• Markdown MathJax equations in Asian text didn't render properly. Fixed.
• Indexed Google Drive documents couldn't be opened via the Data > Open With command. Fixed.
• Links using Apple Notes' URL scheme couldn't be clicked in documents. Fixed.
• For RSS and Atom feeds with multiple authors, only the first was used. Fixed.
• An open Quick Look panel didn't update when changing the selection of the view/edit pane when General > Interface > Preview group content was enabled. Fixed.
• In the item list in List view, the Tags column didn't display autocompletion results. Fixed.
• The tooltip shown when hovering over items with truncated names didn't fully display the name if the item also had property icons. Fixed.
• Groups with a custom icon and applied color displayed differently in different parts of the interface, e.g., the item list and context menu. Fixed.
• In the Navigate sidebar, you could only drag and drop multiple favorites, smart groups, or smart rules by the icon, not the name. Fixed.
• The First, Previous, Next, and Last commands in the Go menu couldn't be used in toolbar search results. Fixed.
• The search field in the context menu and destination popups, e.g., in the Sorter, didn't sort the results by relevance. Fixed.
• In a multi-monitor setup with vertically positioned screens, the Sorter was only partially visible as a docked tab. Fixed.
• The item count in the Database Properties window incorrectly included saved versions. Fixed.
• The internal versions group was incorrectly made visible in the Sorter. Fixed.
• The Spotlight index for a database could incorrectly contain document versions. Fixed, though rebuilding the Spotlight index via Database Properties may be needed.
• .strings files weren't indexed correctly, potentially affecting versioning of AppleScript modified files. Fixed.
• The Scan Text smart rule action didn't ignore Markdown metadata if there was no other content in the document. Fixed.
• If the DEVONthink helper background task crashed, other background tasks, e.g., running scripts, could be blocked for a period of time. Fixed.

• Pro — Processing multiple documents via the Chat assistant could sometimes crash the app. Fixed.
• Pro — Timestamps weren't properly restored when revisiting the Chat assistant. Fixed.
• Pro — Replies in the Chat assistant could contain improperly escaped Markdown. Fixed.
• Pro — The (AI) suffix on an AI-generated document names could be misplaced if the name already had a file extension. Fixed.
• Pro — Claude 3 Haiku responses in the Help assistant often included invalid links. Fixed.
• Pro — Switching to Mistral in the Help assistant and resending the last chat message failed. Fixed.
• Pro — AI-generated web searches containing multiple site: operators didn't work correctly. Fixed.
• Pro — Item links or WikiLinks to groups weren't supported in the AI > Graph inspector. Fixed.
• Pro — The AI > Graph inspector could incorrectly display a connection line to a hidden (disabled) node. Fixed.
• Pro — In the popup of URL fields in the Info > Data inspector and the form view of sheets, the Paste and Remove commands neither saved changes nor updated the layout. Fixed.
• Pro — In the Import > Emails sidebar, nested iCloud mailboxes were duplicated at the account's top-level. Fixed.
• Pro — Search queries using date-based custom metadata weren't always parsed correctly. Fixed.
• Pro — Some item properties, e.g., the URL or UUID, couldn't be retrieved by AI in the Chat assistant, smart actions, or the get chat response for AppleScript command. Fixed.
• Pro — The prompt for Scripts > Chat > Geolocate Image was incorrect. Fixed.
• Pro — The get chat response for message… as HTML AppleScript command could return incorrectly mapped links. Fixed.
• Pro — Gemini 2.5 Pro didn't work with the get chat response for prompt AppleScript command when thinking was disabled. Fixed.

• Server — The icons in the Server > Users settings didn't look correctly in dark mode. Fixed.

• Pro — Removed Data > New from Template > AI > Script Assistant as it has been replaced by the Script Editor and Assistant.
• Pro — Removed support for Google's Imagen 3 (Fast) AI model.
• Pro — Removed support for Anthropic's Claude 3.7 Sonnet and Claude 4 Opus AI models.