Get ready for
UPDATE
Private GPT v1.5 is here!
Smarter, sharper, and more adaptable than ever
We’re excited to announce the release of Private GPT Version 1.5, our most advanced version to date. This release combines deep technical improvements with powerful new features. Designed to deliver faster, more accurate, and more flexible AI support, fully under your control.
FEATURES
What’s new in 1.5?
Private GPT 1.5 is here: Bringing a new level of customization, performance, and usability. With Scenarios 2.0, smarter document processing, and the powerful
Mistral Small model at its core, this release is packed with upgrades that make Private GPT more flexible, more accurate, and easier to use than ever before.
Scenarios 2.0: Total prompt Customization
Scenarios just got a major upgrade: with Scenarios 2.0, you now have full control over prompt content and model behavior.
- Customize system prompts with variables like cutoff_date or user_role
- Adjust key parameters such as creativity, number of chunks, or temperature
- Scenarios are now accessible and configurable via API for easy integration
Whether you’re building a coding assistant, a compliance bot, or a customer service tool: Scenarios 2.0 adapts to your use case.
Smarter Document Parsing & File Support
Our improved parsing pipeline delivers higher-quality input for the LLM:
- Better recognition of structure
(headings, paragraphs, lists) - More accurate and faster OCR
- Improved table extraction
- New file types supported: HTML and XML
- Includes a backup parser to handle problematic documents reliably
Hybrid RAG: Smarter Context Handling
Hybrid RAG enhances retrieval by intelligently choosing full documents (instead of chunks) when they fit the model’s input window, improving response quality without sacrificing performance.
A Fresh Look: Improved User Interface
Private GPT has a new design that not only looks cleaner, but also improves day-to-day usability.
- Refreshed navigation and sidebar layout for better orientation
- Simplified and optimized chat experience
- Fresh colors and modern UI elements for an improved user experience
Improved Prompt Templates
We’ve refined how prompts are structured to improve consistency and response quality.
- Prompts now include system details like current date and model cutoff
- Clearer guidelines to prevent
hallucinations and improve tone - Enhanced instructions for quoting, source differentiation, and ambiguity handling
- Easily override with your own templates via Scenarios 2.0
Mistral Small: Faster, Smarter Model
Private GPT 1.5 now runs on Mistral Small 3.1, a highly capable open-source LLM with:
- Strong multilingual capabilities
- Fast response times
- Better instruction following and context handling
- Optimized for hybrid RAG and new prompt templates
More power, less waiting
RAG History: Context that Follows the Conversation
Our new RAG History Scenario (beta) enables the system to understand follow-up questions across multiple chat turns.
- Automatically rewrites or decomposes user prompts into optimized search queries
- Filters out irrelevant instructions before search
- Pulls relevant context from both the knowledge base and the ongoing chat
- Ideal for natural, multi-turn conversations in research, support, and internal knowledge use cases
Changes & Improvements
- Libre Translate no longer needed thanks to the enhanced language understanding of Mistral Small
- Single Session Chat now always uses full-document context when size allows (up to 65k tokens ≈ 50-60 pages)
- Default parameters for Knowledge Base Chat updated: Chunks increased from 5 → 10 and Similarity threshold lowered for the better recall (0.01 → 0.001)
TLDR;
Why upgrade?
Private GPT 1.5 delivers the performance, flexibility, and control your teams need: whether you're scaling LLMpowered search across your company, building internal tools, or improving access to enterprise knowledge. If you're ready to upgrade or want to learn more, just reach out. The new version is ready for you.
DOKUMENTATION
Want to Learn More?
With Private GPT 1.5, you get a faster, smarter, and more adaptable platform, whether you’re building internal assistants, exploring documents, or scaling secure AI across your organization.
Want to learn more? Visit our Public Documentation Center for detailed guides, manuals, and updates.
As always, thank you for your continued collaboration and support! If you need further information, our team is happy to walk you through the update, just reach out to your project contact at info@pgpt.io. Thank you for being part of this journey with us. Warm regards, The Private GPT Team