# Manticore Search 13.2.3: Faster Vector Search, Lower RAM Usage, and Better Compatibility

Manticore Search 13.2.3 introduces vector search quantization with oversampling and rescoring for faster indexing and lower memory usage, plus key compatibility improvements

We're excited to release [Manticore Search 13.2.3](/install/), the result of our work throughout **June 2025**. This version focuses on making **Vector Search faster, more memory-efficient, and better suited for large-scale data**, while also improving compatibility and stability.

---

## Vector Search: Now Smarter and More Efficient

Since this release, Manticore Search supports [quantized KNN vector search](https://manual.manticoresearch.com/Searching/KNN), giving you:

- **Much faster indexing** of float vector data  
- **Significantly lower RAM usage** due to quantization  
- **Nearly the same accuracy** thanks to oversampling and rescoring  

This is especially important when working with **large datasets** — for example in image search, recommendation systems, or AI-powered search — where indexing performance and memory efficiency are crucial.

---

## Breaking Changes — What You Need to Know

This release includes some internal changes that may affect compatibility if you're using `float_vector` attributes:

- If you **don’t use `float_vector`**, you’re good — **no action needed**.
- If you **do use `float_vector`**, your existing data will be read just fine by the new version. However:
  - If you make changes to your data (e.g. insert or update documents), you may **not be able to downgrade** to an older version afterward.
- The internal API between **Manticore Search** and the **Manticore Columnar Library (MCL)** has changed.  
  If you follow the [official installation instructions](/install/), both packages will be updated automatically — no extra steps needed.

Relevant changes:
- *v13.0.0*: Updated KNN API to allow empty `float_vector` values  
- *v12.0.0*: Fixed incorrect row ID mapping during KNN index training and building  
- *v11.0.0*: Added quantization, oversampling, and rescoring  

---

## Improvements and Fixes

This release also includes:

- Fixes for compatibility with tools like `mysqldump`
- Improved stability in Fuzzy Search and Vector Search queries
- RHEL 10 support
- Better logging and container behavior

We’ve made **over 20 bug fixes** in total. For full details, see the [Changelog](https://manual.manticoresearch.com/Changelog).

---

## Ready to Upgrade?

To benefit from the latest improvements, just follow the [installation instructions](/install/) — both Manticore Search and MCL will be updated automatically.

Need help or want to share feedback?

- Join us on [Slack](https://slack.manticoresearch.com)
- Visit the [Forum](https://forum.manticoresearch.com)
- Report issues or suggest features on [GitHub](https://github.com/manticoresoftware/manticoresearch/issues)
- Or email us at `contact@manticoresearch.com`