# Training a model-based RL agent in Ray

**URL:** <https://discuss.ray.io/t/training-a-model-based-rl-agent-in-ray/15411>\
**Category:** Configure Algorithm, Training, Evaluation, Scaling\
**Created:** [July 31, 2024, 9:07pm UTC](https://discuss.ray.io/t/training-a-model-based-rl-agent-in-ray/15411 "2024-07-31T21:07:37Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![kapibarek](https://sea2.discourse-cdn.com/flex020/user_avatar/discuss.ray.io/kapibarek/32/5698_2.png) [@kapibarek](https://discuss.ray.io/u/kapibarek)\
**Post date:** [July 31, 2024, 9:07pm UTC](https://discuss.ray.io/t/training-a-model-based-rl-agent-in-ray/15411/1 "2024-07-31T21:07:38Z")

</div>

As of the latest release, there is only one model-based algorithm which is DreamerV3.  
I’ve been working with other model-based RL libraries out there but most of the time, the code is either extremely buggy or very difficult to test for basic text-based examples outside the MuJoCo domain. What would be good starting point for someone interested in benchmarking model-based RL algos and incorporating them into Ray?