Unique AI Release Notes: March 2026

Blog Author
by Charlotte Rudigier
Apr 9, 2026
featured image

The 2026.12 and 2026.14 releases bring meaningful improvements across your daily workflows, from a more intuitive chat experience and smarter code execution outputs, to expanded web search capabilities and a more flexible agentic table. This month's updates are all about making Unique AI faster, clearer, and easier to act on.

 

A Fresher, Smarter Chat Experience (BETA)

 

Chat Experience (2)

 

The chat interface has been refreshed to make conversations cleaner and easier to follow:

  • Clearer layout: your messages appear as bubbles on the right, assistant replies on the left, with a tidy question-and-answer flow that works great on mobile.
  • Sources at a glance: files and sources appear as small badges directly below each answer, keeping everything in context.
  • Inline references: references are displayed beneath answers and can be expanded to see the full list at a glance.

 

Inline references

 

 

  • Grouped actions: copy, feedback, and other actions are neatly combined in one place per message
  • One-click table copy: tables include a handy copy button so you can grab data instantly
  • Easier follow-up suggestions: next question suggestions are displayed in a cleaner list layout, easy to scan and tap

 

follow-up suggestions

 

  • Smarter hallucination warning: a discreet floating indicator appears only when a potential inaccuracy is detected, colour-coded by severity so you're informed without being distracted

 

hallucination warning

 

Smarter Knowledge Base & MCP Server Management

  • Usage metrics: activity such as document uploads, searches, scope creation & MCP server usage is now tracked and rolled up into daily, weekly, and monthly metrics per organization.
  • Configurable ingestion limits: Excel and CSV row and column limits can now be set directly in the settings interface, no API calls needed

 

Excel (2)

 

  • Cleaner file selection: files that failed ingestion no longer appear in the context selection window, so you only see content that's ready to use

Bing Web Search Is Now Generally Available

 

Your team can now rely on Bing-powered web search as a production-ready feature. The Bing Grounded Web Search capability has graduated from Beta to General Availability, making it a fully supported alternative to Google Custom Search, which Google is discontinuing.

If your organization is on Single-Tenant SaaS or Self-Hosting, reach out to your point of contact to set up or migrate to Bing search.

 

Code Execution Gets Much More Useful (BETA)

 

When the assistant produces output through code execution, it now appears as interactive cards directly inside the chat:

  • Charts & images: rendered inline with a clear title and the code that created them
  • Tables: displayed as styled, scrollable tables ready to read or copy
  • Files: PDFs, spreadsheets, and more available with a single-click download

 

download

 

  • Faster downloads: generated files are now significantly quicker to retrieve

  • Code Execution has moved to a new "Coding Capabilities" section in Space Admin → Sources & Tools. It runs automatically and is hidden from end users.

 

Code Execution

 

Deep Research Reports Can Now Be Exported

 

Deep Research reports are no longer confined to the chat window. You can now export them directly, making it easy to share findings with colleagues or archive them for future reference.

 

Agentic Table: More Control, Cleaner Exports

 

  • Insert rows: add a new row before or after any selected row without reorganising your sheet manually

 

Insert rows


  • Cleaner DOCX exports: when citations are stripped, the References section is also removed automatically, delivering a clean, ready-to-share document
  • Shared sheet safeguard: non-owners can no longer accidentally trigger file upload actions that would cause errors

 

Unique AI

 

Enhanced Execution Time Logging for Agent Workflows: Debug info now includes a new execution_time section with detailed timing for each agent workflow step — covering planning, tool usage, evaluation, and post-processing,  making it easier to spot performance bottlenecks.

 

More Model Choice, More Power

 

  • GPT-5.4 & GPT-5.4 Pro — now available in Unique AI for global deployments, appearing in the model selection dropdown where eligible (European zone support coming later)
  • Mistral models — the Mistral family is now available for tool selection, giving administrators more flexibility over which model powers specific tasks

These releases reflect our ongoing commitment to making Unique AI a more capable and seamless tool for your team. As always, reach out to your account team or visit Unique Academy if you'd like to explore any of these features in more depth.