2026 Release Notes
November
November 10, 2026 – v2.18.0-LC0.8.8
- Updated chat platform: Upgrades LibreChat to v0.8.8 while keeping the approved features and interface controls.
- Claude Auto: Adds an Auto option that selects a cost-appropriate Claude model for your request. You can still choose a specific model, and existing defaults and usage limits remain unchanged.
- Memories enabled: Activates Memories in production to support a more personalized chat experience.
October
October 20, 2026 – v2.17.0-LC0.8.7
Model Updates
- Added Claude Haiku 5.5. This model with replace Claude Haiku 4.5.
- Added Claude Opus 4.6. This model with replace Claude Opus 4.5.
- Added GPT-6 Luna.
- GPT-6 Luna replaces GPT-5.4 mini as the default model in HopGPT.
Model Retirements
- Claude Opus 4.5, Claude Haiku 4.5, o3, & o3-mini
Improvements
- Response feedback: Enabled LibreChat’s thumbs-up and thumbs-down feedback feature. Users can select a feedback reason and add custom text for negative feedback.
October 13, 2026 – v2.16.0-LC0.8.7
Model Updates
- Removed the 50,000-token context cap for GPT 5.5, GPT 5.6, Terra, and Luna. These models now support longer conversations and larger inputs.
- Added Claude Haiku 5.5 for API Users
Safety and Compliance
- Added a caution notice on pilot agents, warning against entering sensitive or confidential information in agents and explains that agent owners can view all inputs and outputs for agents they created.
User Experience
- Introduced a quarterly feedback survey to collect user input.
Platform Updates
- Upgraded LiteLLM to version 1.102.1.
September
September 29, 2026 – v0.8.7
New Features & Enhancements
- Added Chat Projects for organizing related conversations.
- Added a context-usage gauge showing how much of a model’s context a conversation is using.
- Added Quote Excerpts so you can add to your next prompt as quoted context.
- Redesigned the Settings experience.
- Enhanced retention processing to support multiple HopGPT data stores and Chat Projects.
- Improved secure management of database credentials using Terraform-managed parameters.
- Expanded Bedrock Guardrails threat detection.
New Models for API users
- Added Opus 5.5, GPT-6 Astra, GPT-6 Sol, GPT-6.1 Sol, GPT-6 Luna, & Claude Sonnet 5.5
Bug Fixes
- Fixed Bedrock agent-to-agent handoff failures when requests do not include a tools parameter.
Updates
- Updated the HopGPT Terms of Use and support-site language.
September 8, 2026 – v2.14.0-LC0.8.5
New Models
- Three new models are available: Claude Sonnet 4.6, GPT 5.6 Terra and GPT 5.6 Luna, each with a 50K context window.
Retired Models
- Claude Sonnet 4.5 has been retired. Agents and API users should move to Sonnet 4.6 or to other models.
- Claude Sonnet 4 has been retired.
Bug Fixes
- Bedrock Claude models now handle the verbosity parameter without errors.
August
August 25, 2026 – v2.13.0-LC0.8.5
Bug Fixes
- Streaming API Issue Resolved: Fixed an issue where the system was forcing streaming responses for all requests, even when clients requested standard responses. Non-streaming API requests now correctly return a single JSON response as expected.
- Performance Improvement: Resolved a bug that was causing unnecessary database writes when the transaction feature was disabled, which could slow down system performance. The system now properly respects the transaction configuration setting.
Updates
- Terms of Service Updated: The Terms of Service have been updated with new legal language. You can review the updated terms on the Support site and in the in-app Terms modal. The “Last updated” date has been refreshed to reflect these changes.
New Features & Enhancements
- Claude Opus 5 support: Claude Opus 5 is now available for API users.
- File upload errors restored: You’ll now see clear error messages again when a file is too large or you don’t have permission to upload it, instead of a silent failure.
August 4, 2026 – v2.11.0-LC0.8.5
New Features & Enhancements
- Claude Opus 5 support: Claude Opus 5 is now available for API users.
- File upload errors restored: You’ll now see clear error messages again when a file is too large or you don’t have permission to upload it, instead of a silent failure.
July
July 21, 2026 – v2.10.0-LC0.8.5
New Models
- GPT-5.6 available for API Key users
- Claude Sonnet 5 available for API Key and HopGPT users
Security & Access Control
- Agent knowledge base protection: Users can no longer download knowledge base files attached to an agent
June
June 30, 2026 – v2.9.0-LC0.8.5
New Features & Enhancements
- Claude Opus 4.8 and Claude Opus 4.8 cached are now available for API Key users.
- Added visualization support for reviewing and investigating file deletion events in agents.
Bug Fixes
- Fixed an issue where pinned Anthropic models displayed internal model names instead of user-facing model labels.
June 11, 2026 – v2.8.0-LC0.8.5
New Features & Enhancements
- Refreshed UI: All controls have been moved to the left panel for improved user experience
- Compliance Update: Privacy policy link updated per JHU Office of Compliance requirements
May
May 19, 2026 – v2.7.0-LC0.8.2
New Features & Enhancements
- New Models: gpt-5.4-mini, gpt-5.4-nano, gpt-5.4, and gpt-5.5
- Retired Models: gpt-4.1, gpt-5, gpt-5-mini, gpt-5.1, and gpt-5.2
Bug Fixes
- No bug fixes in this release
Security & Infrastructure
- Addressed a node-level vulnerability to improve platform security
- Expanded internal network connectivity to improve service reliability
- Upgraded core proxy infrastructure to the latest version
- Implemented automated cleanup for internal usage logs to maintain platform performance
Agents
- New HopGPT Support Agent
May 05, 2026 – v2.6.0-LC0.8.2
New Features & Enhancements
- Agent Builder (pilot): Released Agent Builder, currently available to pilot users only.
- Updated the Claude Sonnet context window description to accurately reflect the 50K token limit.
Bug Fixes
- Fixed an issue causing users to be unexpectedly logged out after around 15 minutes of inactivity.
- Improved performance for OpenID users who were experiencing slow load times at login.
- Resolved an issue where large data purge operations would time out before completing.
April
April 21, 2026 – v2.5.0-LC0.8.2
Enhancements
- Increased maximum context tokens for Opus and Sonnet models to 50K, enabling longer and more complex interactions.
Bug Fixes
- Resolved a bug with Vector DB purge (PURGE) to ensure data is properly cleared when expected.
- Fixed an issue where new chats in production did not immediately appear in the left sidebar without a page refresh.
- Implemented a potential fix for a Redis-related issue to improve overall system stability.
April 14, 2026 – v2.4.0-LC0.8.2
Enhancements
- API Reliability – We’ve improved our system’s timeout settings to provide a more stable experience and reduce connection errors.
Bug Fixes
- Vector Database Maintenance – Fixed an issue to ensure data is properly removed from the database when needed, maintaining optimal system performance.
March
March 31, 2026 – v2.3.0-LC0.8.2
New Features & Updates
- LibreChat 0.8.2 – This version brings some new features to the HopGPT user interface:
- Additional description on model hover within the model selection menu
- Model pinning to enable quick access to your favorite models
- Maximum Context Window Update – Limits Opus 4.5 and Sonnet 4.5 to 30K tokens
Enhancements
- LibreChat 0.8.2 – Additional enhancements include:
- Removal of some custom styling to improve accessibility
- Reintroduces multi-conversation mode to provide side-by-side model response
February
February 23, 2026 – v2.2.0-LC0.8.0
New Features & Updates
- LC Environment Variables for HopGPT – Configured environment variables for improved functionality
- Maximum Context Window Update – Claude Opus 4.5 context window limited to 50K tokens
- GPT-5.1 and GPT-5.2 Support – Added compatibility with the latest GPT model iterations
- Model Retirement – Deprecated GPT-4o/4o mini models
February 5, 2026 – v2.1.0-LC0.8.0
Enhancements
- Updated GPT-5 Reasoning Effort setting to Minimal to improve average response time. The model will output very few or no reasoning tokens to improve response speed.
- Upgraded to Next-Gen v1 Azure Open AI API
Bug Fixes
- Fixed Styling Issues in Light Mode
- Fixed Link to TOS
January
January 20, 2026 – v2.0.1-LCv0.8.0
Model Updates
- Deploy o3 DataZone models to HopGPT
Enhancements
- Scientific Notation Display: Values are now presented in scientific notation for improved readability, especially for larger numbers.
- Upgrade LiteLLM to v1.80.11-stable
Bug Fixes
- Accurate LLM Cost Display: The fix resolves inconsistencies in the displayed LLM costs, ensuring users see the correct pricing.
January 14, 2026 – v2.0.1-LCv0.8.0
Model Updates
- Deploy GPT-5 DataZone Models to HopGPT
Enhancements
- Upgrade LiteLLM to v1.80.8-stable
2025 Release Notes
December
December 22, 2025 – v2.0.1-LCv0.8.0
Model Updates
- Released new Anthropic models Sonnet 4.5, Opus 4.5, and Haiku 4.5.
- Retired Anthropic Opus 4 and Sonnet 4 models.
December 16, 2025 – v2.0.0-LCv0.8.0
New Features
- Display the installed LibreChat version in the application footer.
- Added a banner notification to communicate platform upgrades to HopGPT users.
Bug Fixes
- Resolved multiple styling issues across the application.
Enhancements
- Upgraded HopGPT to LibreChat v0.8.0.
- Updated LibreChat Redis secrets to support v0.8.0.
- Updated menus to align with LibreChat v0.8.0. Models are now organized by provider.
- Updated RAG processing so that usernames are no longer sent with requests.
Known Issues
- Some new LibreChat v0.8.0 components remain disabled while further evaluation and testing are completed.
November
November 18th, 2025 – v1.1.3-LC0.7.8
- Users can now upload documents up to 12 MB or more, depending on file content, with a hard limit of 50 MB for text files.
- File Uploads: Implemented a heartbeat mechanism and switched to using Server-Sent Events (SSE) during file uploads to prevent idle timeouts.
- Conversation Imports: Introduced a file size limit when importing conversations via the Settings > Data Controls feature to prevent system crashes. This limit is configurable and is currently set to 250MB.
- Updated production to point to the new URL for the website support: hopgpt.it.jh.edu
October
October 16th, 2025 – v1.1.2-LC0.7.8
- Restored Preset Functionality: Presets are now re-enabled and available in the user interface
- Branding Updates: Update the login page from Hopkin AI Lab to HopGPT
October 14th, 2025 – v1.1.1-LC0.7.8
- Domain changes: Update Production Environment Domain Name to chat.ai.jh.edu
- Removal of multi-chat feature: This feature is not working when LLMs from different providers are selected. Therefore, we have decided to remove it from our platform.
- Rebranding changes: As part of the name change, we have updated some text and replaced all references from Hopkins AI Lab to HopGPT.
September
September 24, 2025 – v1.1.0-LC0.7.8
- Upgrade LiteLLM
- Enable GPT-5 and GPT-5-mini
July
July 29, 2025
- Configured LibreChat and LiteLLM to handle different configurations for dev and prod.
- Enabled Redis-based caching to improve chat performance and reduce response latency.
- Implemented automated CI/CD pipelines to streamline and standardize application deployments across environments.
- Bug fix to handled the 100-second Cloudflare timeout by improving error handling for large file uploads when not on VPN.
- Upgraded RAG to a newer version (v0.5).
- Applied CPU and memory limits/requests to application pods for more efficient resource management and stability.
- Enforced consistent file size limits across TXT, LOG, and CSV file type uploads.
- Updated the Dev DB password for improved security.
July 1, 2025
- Enabled Claude 4 Sonnet and Opus from Bedrock
- A new scheduled cron job has been implemented to automatically purge data older than 30 days in LibreChat and ~6 moths in Langfuse, supporting our ongoing data retention and compliance practices.
- A fix for cached load balancer IP Updates that resolves an issue where outdated IP addresses from the load balancer were being cached.
- Implemented error handling for large file uploads
June
June 9, 2025
- VPN requirement dropped; external access now supported.
May
May 29, 2025
- Update Azure model cost configuration in LiteLLM
- Fixed bug with requests to root being cached
- Update Model Welcome Message: Data Retention Policy
May 27, 2025
- Enabled LLama 4 Maverick, LLama 3.3 Instruct, GPT-4.1 and o3
- Upgrade LibreChat to v0.7.8
- Enabled sticky sessions to address LC scaling issue
- Updated favicon to JH asset
- Removed Claude 3.5 Sonnet from AI Lab
- Fixed Gray text / icons on dark background buttons
- Dynamically Retrieve Mongo CA Certificate at Build Time
- Fixed Styling Issues
April
April 23, 2025
- Implementation of Redis
- UI adjustments
- Bug fix: None Selected
March
March 31, 2025
- Update Model intros
- New Term of Service
- Updated Help & FAQ
- New Azure oai instances
- Scale production
March 20, 2025
- Upgrade to LibreChat 0.7.7
- Upgrade LiteLLM to v1.61.20-stable