# 3.13 Launched 🚀

**URL:** <https://forum.rasa.com/t/3-13-launched/67497>\
**Category:** Announcements\
**Tags:** rasa\
**Created:** [July 22, 2025, 7:38am UTC](https://forum.rasa.com/t/3-13-launched/67497 "2025-07-22T07:38:06Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![Lauren-Goerz](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/lauren-goerz/32/20242_2.png) [@Lauren-Goerz](https://forum.rasa.com/u/Lauren-Goerz)\
**Post date:** [July 22, 2025, 7:38am UTC](https://forum.rasa.com/t/3-13-launched/67497/1 "2025-07-22T07:38:06Z")

</div>

🚀 **Rasa Pro 3.13 is here!** 🚀

We’re excited to share the latest updates in Rasa Pro 3.13 with powerful new features for managing sensitive data, refining enterprise search, and enhancing voice experiences. Here’s what’s new:

🔐 **PII Management in CALM**

Define and manage [personally identifiable information](https://rasa.com/docs/reference/config/pii-management/overview/) directly in CALM. Control what data is marked sensitive, how it’s handled in logs and analytics, and stay audit-ready.

🔍 **Smarter Enterprise Search Triggering**

Improved performance for detecting when to trigger the enterprise search policy via the `SearchReadyLLMCommandGenerator`. You should see fewer false positives and more relevant answers. Read more about it [here](https://rasa.com/docs/reference/config/components/llm-command-generators/#searchreadyllmcommandgenerator).

🗣 **Voice: Global Silence Timeout**

Set a [default silence timeout](https://rasa.com/docs/reference/config/overview/#silence-timeout-handling) for your voice assistant, and override it in specific Collect steps where users may need more time (e.g. entering a phone number).

🔁 **Voice: Repeat Last Turn**

Users can now ask the assistant to [repeat](https://rasa.com/docs/reference/primitives/default-actions/#action_repeat_bot_messages) the _entire_ last turn (not just the last message!), making it easier to stay on track in noisy or interrupted conversations.

🐞 **Bug Fixes & Improvements**

As always, we’ve shipped a range of performance enhancements and improvements under the hood. Check out the [changelog for more information](https://rasa.com/docs/reference/changelogs/rasa-pro-changelog#3130---2025-07-07).

---

<div class="post-metadata">

**Author:** ![cycywanyeki20](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/cycywanyeki20/32/24287_2.png) [@cycywanyeki20](https://forum.rasa.com/u/cycywanyeki20)\
**Post date:** [July 22, 2025, 4:45pm UTC](https://forum.rasa.com/t/3-13-launched/67497/2 "2025-07-22T16:45:38Z")

</div>

Hello Lauren, I have noticed that Rasa 3.13 comes with a CompactLLMCommandGenerator in the config file while the previous version had SingleStepLLMCommandGenerator. What is the impact of this change?

---

<div class="post-metadata">

**Author:** ![Lauren-Goerz](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/lauren-goerz/32/20242_2.png) [@Lauren-Goerz](https://forum.rasa.com/u/Lauren-Goerz)\
**Post date:** [July 23, 2025, 1:21pm UTC](https://forum.rasa.com/t/3-13-launched/67497/3 "2025-07-23T13:21:09Z")

</div>

Hi @cycywanyeki20,

Great to hear from you again 🙂

The compactLLMCommandgenerator was first introduced in 3.12 and includes a prompt template that is optimized for the LLM models `gpt-4o` and `claude-3-5`. GPT-4o is currently the default model if you are using the developer edition.

As models improve and develop, we will need to keep updating and optimizing the command generator on our side to ensure the best Rasa performance for various language models. Today, if you look at our [command generators page in the docs](https://rasa.com/docs/reference/config/components/llm-command-generators/#compactllmcommandgenerator), you can see the command generator prompt template has been optimized for GPT-4o and Claude 3.5 Sonnet.