# Ray 2.7 Released. Check it out!

**URL:** <https://discuss.ray.io/t/ray-2-7-released-check-it-out/12170>\
**Category:** Announcements\
**Created:** [September 18, 2023, 4:28pm UTC](https://discuss.ray.io/t/ray-2-7-released-check-it-out/12170 "2023-09-18T16:28:10Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Jules\_Damji](https://sea2.discourse-cdn.com/flex020/user_avatar/discuss.ray.io/jules_damji/32/4058_2.png) [@Jules\_Damji](https://discuss.ray.io/u/Jules_Damji)\
**Post date:** [September 18, 2023, 4:28pm UTC](https://discuss.ray.io/t/ray-2-7-released-check-it-out/12170/1 "2023-09-18T16:28:10Z")

</div>

Another milestone release for Ray, LLMs, and distributed training.  
Ray 2.7 features major stability improvements for Ray AI Libraries, KubeRay and introduces RayLLM, include GA for Ray Train:  
🚄Simplified APIs in Ray Train for general availability  
💪Stabilized and enhanced Ray Serve and KubeRay for general availability  
🧠 Introduced RayLLM – LLM model serving on Ray Serve  
💪Stabilized Ray Data for an upcoming general availability  
📊:woman-running:Added initial Accelerator support for TPUs, Trainium, and Inferentia

Have read, and try it out. Let us know.

> **[Ray 2.7 features major stability improvements to Ray AI Libraries and KubeRay...](https://www.anyscale.com/blog/ray-2-7-features-major-stability-improvements-to-ray-ai-libraries-and-kuberay-introduces-rayllm)**
>
> From the creators of Ray, Anyscale is a framework for building machine learning applications at any scale originating from the UC Berkeley RISELab.