Changelog
Follow up on the latest improvements and updates.
RSS
new
improved
fixed
Enhanced Text and Math Formula Rendering + Improvements
New
- Support for math formulas and equation rendering.

- Better table rendering.

- New thinking token rendering UI.

- Added support for emojis 😃

Improved
- Better handling of MFA factors, ability to regenerate codes and manual code entry in authenticator app as alternative.
- Overall chat experience update including user messages sending off screen, general formatting, and look/feel.
Fixed
- Added better support for MFA QR codes in dark mode.
new
improved
New AI Phone Agents Interface
New
- New setup wizard and UI with several improvements to make getting started easy.

- Choose response profiles for your AI Agents: Verbose, Concise, Creative

- View call transcripts within the dashboard.

- Load AI Phone Agent Template optimized for specific use cases.

- Limit individual AI Phone Agents to specific users.

- Filter voices by different characteristics in the new interface.

Improved
- Improved AI model selection interface.

- All voices upgraded with both sound quality and realism.
- Various security and speed updates in different parts of the platform.
- Error responses for users attempting to use a model disabled for their organization.
New
- Inline tool calling. Your tool requests from AI no longer happen behind the scenes. See exactly what the AI is doing including search and scrape queries as well as responses from the tools.
Start with a query that includes tools.

The AI will tell you exactly what it is searching, what tools it is using and the responses from the tools.

The AI will continue to use up to four tools for each request. It can reference output from one tool into the next tool call.

Finally, when the AI completes their tool calling it will write you summarize the information.
- Integrations tab. Connect 3rd party integrations to your Hatz AI Account. Navigate to [subdomain]/workspace/integrations to view available integrations.

- Microsoft Outlook integration added. Connect your account by navigating to the integrations tab and completing oAuth. Note: your O365 administrator may need to approve the app.
- Use the outlook integration from inside the secure chat by selecting the "Outlook" button, which will appear after connecting your account. Search emails, list emails, list calendar invites, create email drafts (saved to your drafts folder), send calendar invites, and decline calendar invites.
Note: We have found the best performance with Claude 4 Sonnet.

- Enable organization wide MFA mandate as a General Admin or through the admin dashboard tenant organization settings.

- Enable global MFA mandate for all tenants.

New
- Edit the name and description of drafts for apps and workflows.

- Use images as an input in workflows for steps using LLMs that support vision.

New
- Set default LLM globally. This feature is available in the admin dashboard and will change the default LLM for all tenants that do not have their own default LLM selected. https://admin.hatz.ai/settings/global-settings

- Set the default LLM for a specific tenant from the admin dashboard.

- Set the default LLM as a General Admin.

- Disable LLMs Globally.

- Disable LLMs for specific tenants.

- Disable LLM as General Admin

Fixed
- Fixed a bug causing some users to not be able to save edits to apps and workflows created by another user.
New
New LLMs including Open AI's o3 and o4-mini, Google Gemini 2.5 Pro and Flash, and Llama 4 Maverick and Scout.

- Description from OpenAI: OpenAI o3 is our most powerful reasoning model that pushes the frontier across coding, math, science, visual perception, and more. It sets a new SOTA on benchmarks including Codeforces, SWE-bench (without building a custom model-specific scaffold), and MMMU. It’s ideal for complex queries requiring multi-faceted analysis and whose answers may not be immediately obvious. It performs especially strongly at visual tasks like analyzing images, charts, and graphics. In evaluations by external experts, o3 makes 20 percent fewer major errors than OpenAI o1 on difficult, real-world tasks—especially excelling in areas like programming, business/consulting, and creative ideation. Early testers highlighted its analytical rigor as a thought partner and emphasized its ability to generate and critically evaluate novel hypotheses—particularly within biology, math, and engineering contexts.

- Description from OpenAI: OpenAI o4-mini is a smaller model optimized for fast, cost-efficient reasoning—it achieves remarkable performance for its size and cost, particularly in math, coding, and visual tasks. It is the best-performing benchmarked model on AIME 2024 and 2025. In expert evaluations, it also outperforms its predecessor, o3‑mini, on non-STEM tasks as well as domains like data science. Thanks to its efficiency, o4-mini supports significantly higher usage limits than o3, making it a strong high-volume, high-throughput option for questions that benefit from reasoning.

- Description from Google: Gemini 2.5 Pro is our state-of-the-art thinking model, capable of reasoning over complex problems in code, math, and STEM, as well as analyzing large datasets, codebases, and documents using long context. Gemini 2.5 Pro rate limits are more restricted since it is a preview model.

- Description from Google: Our best model in terms of price-performance, offering well-rounded capabilities. Gemini 2.5 Flash rate limits are more restricted since it is an experimental / preview model.

- Description from Meta: Llama 4 Scout, a 17 billion active parameter model with 16 experts, is the best multimodal model in the world in its class and is more powerful than all previous generation Llama models, while fitting in a single NVIDIA H100 GPU. Additionally, Llama 4 Scout offers an industry-leading context window of 10M and delivers better results than Gemma 3, Gemini 2.0 Flash-Lite, and Mistral 3.1 across a broad range of widely reported benchmarks.

- Description from Meta: Llama 4 Maverick, a 17 billion active parameter model with 128 experts, is the best multimodal model in its class, beating GPT-4o and Gemini 2.0 Flash across a broad range of widely reported benchmarks, while achieving comparable results to the new DeepSeek v3 on reasoning and coding—at less than half the active parameters. Llama 4 Maverick offers a best-in-class performance to cost ratio with an experimental chat version scoring ELO of 1417 on LMArena.
- Display of thinking tokens fro reasoning models. Current supported models include DeepSeek, o3, and o4-mini.

Improved
- Improved performance and speed for DeepSeek.
new
improved
fixed
New LLM Selector, Web Scraping Step, & Various Improvements
New
- New LLM Selector with featured LLMs, model attributes, credit bands, searching, and other improvements.
View Featured LLMs

Sidebar for Viewing All Models

Filter by Developer

Model Attributes
- Model Descriptions
- Capabilities (Low, Medium, High)
- Speed (Slow, Medium, Fast)
- Credit Usage (Lowest, Low, Medium Low, Medium High, High, Premium)
- Vision capabilities (Images)
- Tool usage (AI-Powered Search + Other Tools)

- New model selector and sidebars in workflows and apps


- Scrape websites as a step in workflows. Use inputs, outputs, or text.





Improved
- Autoscroll in chat adjusted to make it easier for users to read outputs as they stream.

- Better responsiveness with the chat sidebar and other chat improvements.
Fixed
- Fixed a bug causing some users to experience stream errors with long outputs lasting longer than two minutes.
new
improved
fixed
Claude 4 Sonnet, Claude 4 Opus & Chat Enhancements
New
- New LLMs: Claude 4 Sonnet and Claude 4 Opus. Claude Opus 4 is the world’s best coding model, with sustained performance on complex, long-running tasks and agent workflows. Claude Sonnet 4 is a significant upgrade to Claude Sonnet 3.7, delivering superior coding and reasoning while responding more precisely to your instructions.

Improved
- Various chat improvements including styling updates to code blocks, text-input auto resizing, and pasting user experience.
Fixed
- Fixed bug causing legacy word files to sometimes fail file upload.
New
- View yourrun history of AI Workflows. You can open the workflow run history for a specific workflow from the run view or view all run history from the "Run History" tab in Workshop. Review credits used for each step, time to run each step, and compare inputs with outputs.




- Generate, view, and manage API keys outside of the customer dashboard. End customers can now generate their own API Keys.

- End customer MFA settings can now be accessed by navigating to Workspace > Account
Improved
- Various improvements to AI-Powered search including the swapping of Brave Search for Google Search. The AI-Powered search button will also now stay clicked throughout a chat.
Fixed
- Various fixes for File Upload. Some users experienced errors when uploading relatively larger files that were still under the file size limit. Improvements made to the copying and pasting of files and Office Document Contents sometimes pasting as images.
New
- Introduced a redesigned chat experience with a new chat sidebar, chat window, and chat input, supporting chat creation, pinning, editing, deletion, and advanced file uploads (including OneDrive integration).
- Added the ability to rename chats with better deletion from the chat sidebar.

- Added a powerful, filterable, and sortable chat history table with multi-select, bulk actions, and improved search and filtering options.

- Integrated AI-assisted HTML editing and suggestions in email template editing.

Improved
- Improved the chat input box to add more spacing and an adjustable height.
- Enhanced chat management with real-time updates, optimistic UI, and animated transitions.
- Improved chat input with file attachment previews and AI-powered search toggle.
- Updated markdown rendering with improved code block copying and syntax highlighting.
Bug Fixes
- Adjusted muted text color for better readability in both light and dark themes.
- Improved error handling and user feedback in file upload and chat streaming features.
Load More
→