Skip to main content
Triton AI Triton AI
  1. TritonAI
  2. TritonGPT
  3. Release Notes
  4. November 24, 2025 Release Details

November 24, 2025 Release Details

Overview

This release refreshes the Chat UI and organizes Documents and Chat Folders into "Projects." It also adds GPT-5.1, Gemini 3 Pro, Deep Research, and new personalization options. See the changes in our interactive button guide.

What's New and Changed

Projects: Your Personal AI Workspace

  • Unified Organization: "My Documents" and "Chat Folders" have been consolidated into a single feature called Projects. This unifies your files and conversation history into one streamlined workspace.
  • Personalized Agents: Specific custom instructions can now be added directly within a Project. This allows you to tailor the behavior of TritonGPT for that specific project, effectively turning it into your own personal agent or specialized assistant for that topic.

Personalization

  • Control Your Experience: Manage your TritonGPT interactions by clicking on your name in the bottom-left corner of the screen.
  • Default Model Selection: Choose your preferred default LLM.
  • Memory & Customization: This new menu enables specific "Personalizations" and "Memory." These preferences are integrated into the system prompt while you use TritonGPT, tailoring the AI's responses to your specific needs and context.
  • Smart Defaults: To get you started, the system automatically populates your Display Name and UCSD Affiliation. However, you retain full control to remove or edit these details to best suit your preferences.

Deep Research

  • Search Capabilities Return: Internet search functionality is back in TritonGPT, allowing for up-to-date information retrieval directly within the platform. Use this feature combined with Deep Research to develop rich documents and analysis that combines information from UC San Diego and public sources.
  • How to Access: Access these features by selecting the Internet Search Assistant located within the Agent Selection Panel.

New Models & Enhanced Privacy Labels

  • New Cloud Capabilities: GPT-5.1 and Gemini 3 Pro are now available, offering access to the latest generation of cloud-based AI reasoning.
  • Secure Local Options: Our existing on-premises models, including GPT OSS and Llama 4 Scout, remain available for use.
  • Privacy Transparency: To help you make informed decisions, models are now clearly labeled as "Cloud" or "Local" directly in the interface.
    • Local identifies a UC-hosted model route. The label does not by itself approve a model for sensitive data; users must follow the data-handling conditions for the service and project.

System Performance & Updates

  • Refreshed Interface: The main chat interface has received a visual update for a cleaner, more intuitive experience. This includes fixes for various visual bugs and improved branding consistency throughout the platform.
  • Job Description Helper: The JDH tool has been upgraded to run on our improved backend infrastructure for better reliability.
  • Better Complex Task Handling: The system can handle longer chains of actions in a multi-step request.
  • Faster Search: Search speeds have been optimized to be faster and even more accurate.