Welcome to Volume 2 of our 2026 product release notes! We've been working hard to bring you performance improvements that add value to your work. Read on to learn about the product enhancements and bug fixes released in February 2026.
Benchling AI
Benchling's artificial intelligence agents and features.
Ask, Deep Research, Compose, and Data Entry Agents are generally available [2026.2.AI.1]
Preview labels have been removed from Ask, Deep Research, Compose, and Data Entry Agents. These features are now generally available.
Custom instructions for Benchling agents [2026.2.AI.2]
Users can now customize Agent behavior by adding custom instructions in the settings menu. Guidelines apply to all Agents including Deep Research, Compose, and Ask, allowing users to specify details like role descriptions, language preferences, or formatting instructions.
Use Ask to get questions answered about Insights dashboards [2026.2.AI.3]
Added a new Ask button to Insights dashboards, allowing users to ask questions about dashboard data.
Ask understands workspace context more intelligently [2026.2.AI.4]
The Ask chat input now displays the active workspace item as a chip, making it clear what context is included. It can also include relevant items (up to 50 per category) from the active workspace when generating responses. When user requests are slightly vague, Ask will make reasonable assumptions rather than always asking clarifying questions.
Ask displays Benchling object chips and searches attachment content [2026.2.AI.5]
Ask now renders object references as interactive, clickable chips in all response types. It can also search within attachment content, including PDFs and documents attached to notebook entries, and clearly cites where information came from such as entry text or a specific file.
Deep Research locates items faster and generates better visualizations [2026.2.AI.6]
Deep Research now uses named entity recognition (NER) to identify Benchling objects mentioned in research prompts, enabling faster lookup of specific entities and IDs. Plot rendering size issues have also been resolved.
Deep Research output and capacity expanded [2026.2.AI.7]
Deep Research now supports longer output responses and handles extended sub-agent conversations automatically. When a conversation approaches the token threshold, the system summarizes and continues in a new thread without losing context. Sub-agents no longer fail due to excessive retries on long answers, and timeout errors on detailed reports are significantly reduced.
Compose improvements: subtemplates, table coordinates, and well plate handling [2026.2.AI.8]
Compose has increased in output capacity and can now reference and search for subtemplates. Table editing is more reliable following a fix for a cell indexing error. When attempting to add a row to a well plate with fixed dimensions a clear error message will now be returned.
SQL Writer capacity expanded [2026.2.AI.9]
SQL Writer now handles warehouses with large schemas and supports double the previous output token limit, enabling more complex queries for advanced analytics without context length errors.
AI models upgraded across all agents [2026.2.AI.10]
Models have been upgraded for Ask, Deep Research, and Compose Agents
Thinking budget extended to all model providers [2026.2.AI.11]
Thinking budget and reasoning support are now available across all model providers, including non-reasoning OpenAI models (e.g., GPT-4.1).
Agent reliability with searching and error handling improvements [2026.2.AI.12]
Fixed issues where LLM requests would fail with unhandled errors when content filtering policies were triggered, and where Ask would fail to analyze oversized search result documents. Sequence search now more accurately incorporates search parameters into queries.
Notebook
Benchling Notebook helps users log their experiments, create protocols, and capture results against their samples.
Entry version resets reflected in audit history [2026.2.N.1]
Entry audit logs now display a specific 'Reset document version' action when reverting to a previous version, making it clear who performed the reset and when. Previously, version resets appeared as generic 'Update content' actions.
In Vivo
Manage in vivo studies with modern software designed for data capture, collaboration, and insights.
Improved sample plan configuration with auto-naming and smart duplication [2026.2.IV.1]
Setting up sample plans is now faster and less error-prone. Sample definition names auto-populate based on your selected sample type, and when you add additional rows, all metadata from the previous definition automatically carries over. This is especially useful when configuring multiple definitions for the same tissue type with different assays—simply duplicate and adjust what's different, rather than re-entering all the details.
Observation labels sync to Benchling Core [2026.2.IV.2]
Observation assay result schemas now include the observation label field when syncing to Benchling Core through webhooks. This aligns observations with samples, enabling the label to appear in Benchling warehouse for AI and insight reports.
Flexible baseline selection for relative calculations [2026.2.IV.3]
When plotting study data post disease induction date, users can now choose whether percentage change and relative calculations use disease induction date or tracking date as the baseline. This gives users precise control to match their protocol requirements—essential for oncology and complex disease models where the comparison baseline differs from the plotting reference. Configure this setting in Treatment Groups, and it automatically updates all graphs and exports.
[Beta] Calculate dose volume in mL [2026.2.IV.4]
Select dosing volume unit to calculate dosing volume in mL rather than uL for large animals.
Contact support to enable
[Beta] Bulk Record Observation [2026.2.IV.5]
Users can now record observations in bulk across multiple animals from the Animals table, whether marking an entire cohort as "Normal" during daily cage-side checks or applying specific observations to filtered groups. This eliminates the need to record observations one animal at a time, dramatically reducing the time spent on routine welfare documentation while maintaining complete audit trails for compliance
Contact support to enable
[Beta] View latest observation directly from the Animals table [2026.2.IV.6]
The Animals table now displays the most recent observation date and recorded observations (up to 3) for each animal at a glance. This gives users immediate visibility into animal health status without navigating away from their primary workspace, helping users quickly identify which animals need attention and accelerate decisions around randomization or veterinary review.
Contact support to enable
Registry and Inventory
The Registry and Inventory combine to form an extremely powerful solution for life sciences R&D samples.
Entity map [2026.2.RI.1]
Entity Map gives scientists a clear, interactive view of how their entities are actually connected in Benchling. This is a new tab on each entity that enables users to visualize relationships between entities in an interactive graph and easily navigate to related entities.
Insights
Insights enables users to query, visualize, and share high quality, structured data that resides in Benchling.
Analysis sidebar displays Regression child views [2026.2.IN.1]
Fixed an issue where analysis regression steps with child views added by users would not appear in the sidebar navigation. The sidebar now correctly displays regression views that contain additional child steps while maintaining the simplified view for standard regressions.
Analysis left sidebar now resizable [2026.2.IN.2]
Users can now resize the left sidebar in analysis views by dragging the separator between the sidebar and main content area. The sidebar width can be adjusted between minimum and maximum bounds to customize the workspace layout.
Worksheet references display as chips in Insights dashboards [2026.2.IN.3]
Worksheet references in Insights dashboards now display as interactive chips showing worksheet names instead of raw API IDs. Clicking a chip opens the corresponding worksheet. CSV exports convert worksheet chips to readable worksheet names automatically.
Custom Views audits [2026.2.IN.4]
Custom Views in Insights Dashboards now has audit log exports
Add to worklist from Dataset [2026.2.IN.5]
Entities in a Dataset column can now be added to a worklist, as long as the column values are entity IDs/chips. The option is available from the Dataset's overflow menu by selecting Add to worklist. The flow follows the same pattern as other "add to worklist" flows in Benchling.
Benchling Connect
Benchling Connect, formerly Lab Automation, enables seamless integration of Benchling to liquid handlers, plate readers, imaging instruments, and more.
Condition replicate mappings column displayed in lab automation run for Results when parallel execution is used (Bioprocess-specific). [2026.2.BC.1]
When a lab automation run schema is configured to map data to a specific Condition - Replicate, Benchling now shows the outcome of that mapping in the Result table that is displayed in the lab automation run inside a parallel execution worksheet. This change is just for Results. A change in two weeks will extend this to Entities and Containers.
Schemas
Schemas are the structural foundation for how your information is categorized in Benchling. They are categories created specifically for your tenant that can represent materials, data, storage methods, and tasks.
Improved error handling for malformed Dropdown updates in V2 API [2026.2.SCH.1]
Fixed an issue where updating dropdown options via the V2 API would return a 500 server error when the request included unsupported properties like archiveRecord. The endpoint now returns a clear 400 bad request error with validation details when malformed data is submitted.
Application Core
To create a seamless Benchling experience, our platform must be consistent, compliant and predictable. The Application Core provides the central capabilities to create a smoother user experience.
Enabled ability to use tenant roles on orgs, teams, and apps [2026.2.AC.1]
The initial release of tenant roles only allowed them to be used on users. This release expands that capability to allow customers to assign tenant roles like User or App Admin to groups of users (orgs or teams) or to Apps, which can reduce the toil of managing tenant-scoped access overall.
New option for not requiring a reason for change when new rows are added to a submitted table [2026.2.AC.2]
If enabled, when a user submits new rows to a previously submitted table, it will not trigger the Data Mods Reason for Change modal. Note that the new row additions will still be audited and that if also editing existing data when submitting new rows will still trigger the Data Mods Reason for Change modal
Contact support to enable
Compliance Controls
With Compliance Controls, gathering required information for compliance is simple, intuitive, and performant. The team's primary focus is on our auditing platform but we also own exporting and archiving.
Large project exports now split automatically [2026.2.CC.1]
Improved large project exports by automatically splitting the export into multiple downloadable parts when needed.
Developer Platform
Our developer platform is fundamental to centralizing and standardizing all your R&D data. Benchling's APIs and Events are built to match the flexibility and speed of modern life science R&D.
V3 REST API open beta [2026.2.DP.1]
The V3 REST APIs, incl. all available endpoints in alpha, beta, and stable are now available in open beta.
Contact support to enable
App Canvas markdown links open in new tabs [2026.2.DP.2]
Links in Canvas Markdown UI Blocks now automatically open in new browser tabs. This is an improvement over the previous behavior where links opened in the same tab, disrupting form submission and other workflows in Benchling.
Structured table names in V2 APIs [2026.2.DP.3]
V2 Entry APIs now return the name attribute for Structured Tables (Results Tables and Lookup Tables). This allows API consumers to display user-friendly table names instead of relying solely on API identifiers when retrieving entry content.
PipeBio
Benchling's cloud-based bioinformatics software for antibody/peptide screening & drug development. Conduct BCR, TCR, VHH, scFv, or peptide sequence analysis and integrate your workflows with Benchling's core platform.
Improved layout for filters [2026.2.PB.1]
An improvement has been made to filters in PipeBio documents and jobs tables. Previously, very long filters would overlap each other. Filters are now placed on a new line in the UI with a horizontal scrollbar so all filters can be easily seen.
Added SQL `LIKE` filters in the jobs table [2026.2.PB.2]
It is now possible to filter jobs in the jobs table using the SQL operator LIKE for comparisons, similar to the functionality in PipeBio documents. For example, to search for jobs having a 1 in the name, a filter such as name LIKE '%1%' can be used.
Multiple charts of the same type [2026.2.PB.3]
It is now possible to create multiple charts of the same type within a PipeBio document using the Sequences > Chart menu or the + icon in the charting sidebar. Chart names can also be edited via right-click.
Bug Fixes
Benchling AI
- Notebook check supports large entries
- LLM chat progress messages display reliably
- AI data entry handles content filtering errors gracefully
- File uploads work at the start of new chats
- Fixed an issue where retry buttons would fail or show confusing error messages in Deep Research and Ask agents
- Deep research credits tracked reliably
- Fixed an issue where some Academic users were blocked from using AI features
- Fixed an issue where Agent selector were not displayed in the correct order
- Fixed an issue where tool errors were that displayed clearly
- Fixed an issue where Deep Research would fail if MCP server encountered an issue
- Fixed an issue where Deep Research would retry queries too many times
- Fixed an issue where Ask wasn't above to respond with enough characters
- Fixed an issue where share link chat included a feedback button
- Fixed an issue where SQL Writer did not reliably generate properly formatted outputs.
- SQL Writer avoids reserved keyword aliases
- PDF attachments display correct image formats
- Deep Research gracefully handles model access errors
Notebook
- Fixed an issue with default formula values being incorrectly updated
- Resizing restored for excerpts and attachments
- Duplicate box option removed from paste menu
- Fixed an issue where exports would be blocked when the creator of the respective entry no longer had READ access
- Fixed where calendar displayed checklists from multiple notebook sections
- Fixed where review lifecycle configuration handles had missing access
- Notebook performance improved when opening steps
- Fixed error blocking registration when users changed from 'Create New Entities' table to a 'Register Existing Entities' table during configuration
- Fixed where Plate and box columns allowed references to multiple items.
- Fixed an issue with default formula values being incorrectly updated
- Pasting box chips into tables works correctly
- Fixed column resizing in structured tables
- Fixed AI chat share links opening from entries
- Fixed an issue where older worksheets weren't included in Relevant Items
- Notebook Entry Metadata Issue Resolution
Molecular Biology
- Fixed overlapping parts bug
- AA sequence multi-FASTA export restored
- Global create menu preserves scroll position
- Performance improvement when switching tabs with alignments
- Oligo duplex schema creation without components
- Ensembl database imports is running reliably
- Performance improvement to alignments
- Fixed where sequence imports validate unsupported characters
Small Molecule
- Fixed incorrect unit validation within structured table cells
In Vivo
- Graph display now matches data table and exports for disease induction plotting
- Multi-row animal selection with Ctrl+Shift+click
Registry and Inventory
- Fixed an issue impacting storable link fields when entering sample barcodes if storable link data was unexpectedly null
- Entry schemas now excluded from validation error filter
- Parent link field updates support large schemas
Workflows
- Entry execution task tables prevent invalid rows
- Fix issue with error previewing imported files in Import Files from Data Catalog
- Fixed headers in request templates and removed request template tab
Insights
- Analysis output column mapping improved accuracy
- Insights queries handle mixed null values
- Fixed issue with Analysis Templates with look up steps being incorrectly configuration migrated
- Dataset analysis dropdown displays correctly on screen
- CSV upload steps support adding downstream steps when upload in progress
- Fix unclear error message when data type mismatch in creating results from analysis template
- Analysis filter options work without Insights Labs
Benchling Connect
- Improved reliability for “Send to Document” step in Analysis
- Fix run schema configuration migrations support in GxP tenants
Schemas
- Fixed where dropdown pages did not load correctly via URL
- Fixed where Schema naming strategies do not update correctly with fieldset/schema config migration
- Fixed Container name template incorrect auditing
Application Core
- Results API provides better error handling for invalid legal text values
- Fixed issue of results table creation permissions for results created from a request/request template
- Fixed an issue where admins couldn't suspend users in some cases if they had warehouse credentials
Developer Platform
- UserId now included in canvas.created spec
- Protein annotation end position bug in SDK
PipeBio
- Cluster multiple samples fix for paired documents
- Amino acid heatmap download fix