Across releases 2026.32, 2026.34 and 2026.36 this month's highlights center on a faster, more consistent Knowledge Base, a more capable Agentic Table, and expanded access to the latest AI models. Here's what's new, and what it means for your day-to-day work.
Unique AI Chat - Smoother, Smarter Everyday
The Model selection in chat is now grouped by provider.
.png?width=2091&height=1104&name=model%20choice%20(1).png)
A "jump to latest message" button now appears whenever you've scrolled away from the bottom of a conversation, so you can catch up in one click.
And your assistant is now aware of every file you upload to a chat, including images and Excel files that aren't ingested into the Knowledge Base, so it can reason about what you've shared even when a file isn't part of your searchable knowledge. If a web search hits a problematic URL, the platform blocks just that link and continues crawling the rest instead of aborting the whole batch, so a single unusable result no longer means a lost answer.
RAG - A Faster, More Consistent Knowledge Base
Files now open with the same preview and navigation everywhere — Knowledge Base, canvas, and chat — so browsing between documents works the same way no matter where you start. You can also drag to resize the file preview pane, giving you more room to read a document while you work.
Search and browsing are noticeably faster too: results stay responsive even when a space is scoped to a large subset of files, and the folder tree loads faster for spaces with dynamic scope rules. When a file does fail to ingest, the status now tells you why (for example, a timeout on an image-heavy document) instead of a generic failure message.
PDF ingestion is also more reliable: table headers are no longer misread from the first row when it isn't one, and, for admins who need the full picture, dropdown form fields can now be ingested with their complete list of choices rather than just the value that was selected.
Mobile users get a smoother experience too: the Internal Search file panel now opens and closes the way you'd expect, and selected tools no longer cover the model picker on small screens.
Agentic Table - A More Capable Agentic Table
Agentic Table cells can now render images, using the same notation already supported in chat messages.
.png?width=2091&height=1104&name=Rendered%20Image%20Table%20(1).png)
The “Can Manage role” is now generally available for share and assign flows.
.png?width=2091&height=1104&name=Share%20Table%20(1).png)
Large sheets also scroll more smoothly, loading rows in batches as you reach the end of what's loaded, and Answer Library entries now carry their client, strategy, and valid-as-of tags through automatically, saving a step every time a strong answer is promoted into your library.
Agentic Table also gains per-row dynamic metadata filters, so each sheet and row, can search a different slice of the Knowledge Base, such as only documents valid as of a given cutoff date.
.png?width=2091&height=1104&name=Table%20Instructions%20(1).png)
AI Models - Expanded Access to the Latest AI Models
Your model options keep growing. Claude Opus 5 and Gemini Flash 3.6 are all now available on the platform. The model dropdown in chat is also now grouped by provider, making it easier to find the model you want. Claude Opus 4.1 was retired on August 5, no action is needed on your end.
For Platform Administrators
A new downloadable per-file ingestion report maps Unique content IDs back to source files, making it far easier to reconcile inventories and troubleshoot failed ingestions. The detailed chat interactions export now includes a fuller set of ROI metrics — cost, token usage, tool and source counts, and response time — for reporting on time saved.
Connectors
New connectors for Outlook Calendar and Backstop are available.
What's Coming Next
Unique AI Conduct is reached its Beta milestone, with development focused on longer-term memory, more proactive assistance, and genuine end-to-end workflows. You'll also be able to turn on Context Memory, so your assistant remembers your working style across conversations, and Shared Chats, for sharing a read-only snapshot of a conversation with colleagues via a link.
On the notification side, informative notifications are getting daily and weekly caps to keep them useful rather than noisy, and a redesigned Knowledge Base UI — with list views, filters, and a detail pane for faster browsing — is moving toward general availability. Our Premium Benchmarking is now in experimental phase, which will enable deeper performance comparisons.
As always, these updates roll out automatically; there's nothing for your team to install. If you'd like a walkthrough of any of these features, or want to enable one that isn't yet active in your workspace, reach out to your customer success contact.
What's New in Unique AI : September 2026
Across releases 2026.32, 2026.34 and 2026.36 this month's highlights center on a faster, more consistent Knowledge Base, a more capable Agentic Table, and expanded access to the latest AI models. Here's what's new, and what it means for your day-to-day work.
Unique AI Chat - Smoother, Smarter Everyday
The Model selection in chat is now grouped by provider.
A "jump to latest message" button now appears whenever you've scrolled away from the bottom of a conversation, so you can catch up in one click.
And your assistant is now aware of every file you upload to a chat, including images and Excel files that aren't ingested into the Knowledge Base, so it can reason about what you've shared even when a file isn't part of your searchable knowledge. If a web search hits a problematic URL, the platform blocks just that link and continues crawling the rest instead of aborting the whole batch, so a single unusable result no longer means a lost answer.
RAG - A Faster, More Consistent Knowledge Base
Files now open with the same preview and navigation everywhere — Knowledge Base, canvas, and chat — so browsing between documents works the same way no matter where you start. You can also drag to resize the file preview pane, giving you more room to read a document while you work.
Search and browsing are noticeably faster too: results stay responsive even when a space is scoped to a large subset of files, and the folder tree loads faster for spaces with dynamic scope rules. When a file does fail to ingest, the status now tells you why (for example, a timeout on an image-heavy document) instead of a generic failure message.
PDF ingestion is also more reliable: table headers are no longer misread from the first row when it isn't one, and, for admins who need the full picture, dropdown form fields can now be ingested with their complete list of choices rather than just the value that was selected.
Mobile users get a smoother experience too: the Internal Search file panel now opens and closes the way you'd expect, and selected tools no longer cover the model picker on small screens.
Agentic Table - A More Capable Agentic Table
Agentic Table cells can now render images, using the same notation already supported in chat messages.
The “Can Manage role” is now generally available for share and assign flows.
Large sheets also scroll more smoothly, loading rows in batches as you reach the end of what's loaded, and Answer Library entries now carry their client, strategy, and valid-as-of tags through automatically, saving a step every time a strong answer is promoted into your library.
Agentic Table also gains per-row dynamic metadata filters, so each sheet and row, can search a different slice of the Knowledge Base, such as only documents valid as of a given cutoff date.
AI Models - Expanded Access to the Latest AI Models
Your model options keep growing. Claude Opus 5 and Gemini Flash 3.6 are all now available on the platform. The model dropdown in chat is also now grouped by provider, making it easier to find the model you want. Claude Opus 4.1 was retired on August 5, no action is needed on your end.
For Platform Administrators
A new downloadable per-file ingestion report maps Unique content IDs back to source files, making it far easier to reconcile inventories and troubleshoot failed ingestions. The detailed chat interactions export now includes a fuller set of ROI metrics — cost, token usage, tool and source counts, and response time — for reporting on time saved.
Connectors
New connectors for Outlook Calendar and Backstop are available.
What's Coming Next
Unique AI Conduct is reached its Beta milestone, with development focused on longer-term memory, more proactive assistance, and genuine end-to-end workflows. You'll also be able to turn on Context Memory, so your assistant remembers your working style across conversations, and Shared Chats, for sharing a read-only snapshot of a conversation with colleagues via a link.
On the notification side, informative notifications are getting daily and weekly caps to keep them useful rather than noisy, and a redesigned Knowledge Base UI — with list views, filters, and a detail pane for faster browsing — is moving toward general availability. Our Premium Benchmarking is now in experimental phase, which will enable deeper performance comparisons.
As always, these updates roll out automatically; there's nothing for your team to install. If you'd like a walkthrough of any of these features, or want to enable one that isn't yet active in your workspace, reach out to your customer success contact.