# Recommendations for LLM nodes sizing
**URL:** https://discuss.elastic.co/t/recommendations-for-llm-nodes-sizing/383813
**Category:** Elasticsearch
**Created:** [December 2, 2025, 1:34pm UTC](https://discuss.elastic.co/t/recommendations-for-llm-nodes-sizing/383813 "2025-12-02T13:34:24Z")
**Posts on this page:** 7
**Page:** 1
### Author:  [@michael-a](https://discuss.elastic.co/u/michael-a)
#### Post date: [December 2, 2025, 1:34pm UTC](https://discuss.elastic.co/t/recommendations-for-llm-nodes-sizing/383813/1 "2025-12-02T13:34:24Z")
We’re about to implement our first LLM node (on Kubernetes) but haven’t been able to find current recommendations when it comes to sizing it. Are there any general rules we can apply based on number of events, log size etc. or do you have any guidelines that we can get hold of?
---
### Author:  [@leandrojmp](https://discuss.elastic.co/u/leandrojmp)
#### Post date: [December 2, 2025, 2:15pm UTC](https://discuss.elastic.co/t/recommendations-for-llm-nodes-sizing/383813/2 "2025-12-02T14:15:41Z")
What do you mean with LLM node?
This is not a node role in Elasticsearch.
Are you talking about a Machine Learning node?
---
### Author:  [@michael-a](https://discuss.elastic.co/u/michael-a)
#### Post date: [December 2, 2025, 2:40pm UTC](https://discuss.elastic.co/t/recommendations-for-llm-nodes-sizing/383813/3 "2025-12-02T14:40:26Z")
Err yeah, my bad - ML node, too late in the day and abbreviation mixup..
---
### Author:  [@leandrojmp](https://discuss.elastic.co/u/leandrojmp)
#### Post date: [December 2, 2025, 2:50pm UTC](https://discuss.elastic.co/t/recommendations-for-llm-nodes-sizing/383813/4 "2025-12-02T14:50:40Z")
The sizing will depend more on the jobs that you are going to run, but the recommendation seems to have a 64 GB node for ML per this [post](https://www.elastic.co/blog/sizing-machine-learning-with-elasticsearch).
Just to confirm, you have a Platinum or Enterprise license, right?
---
### Author:  [@michael-a](https://discuss.elastic.co/u/michael-a)
#### Post date: [December 3, 2025, 6:09am UTC](https://discuss.elastic.co/t/recommendations-for-llm-nodes-sizing/383813/5 "2025-12-03T06:09:00Z")
Thanks @leandrojmp - had seen that post earlier but as it’s more than 8 years old, I assumed it was outdated but if that’s still valid, then that’s what we’ll use. Oh and yes, we’re in the process of upgrading our license from Basic to Enterprise so we’re good to go - at least from a licensing point.
---
### Author:  [@leandrojmp](https://discuss.elastic.co/u/leandrojmp)
#### Post date: [December 3, 2025, 12:17pm UTC](https://discuss.elastic.co/t/recommendations-for-llm-nodes-sizing/383813/6 "2025-12-03T12:17:57Z")
The memory usage will depend on which jobs you have and your data, it is not something that you can easily estimate.
You may start with a smaller node, something like 16GB and increase the memory if you have any issues.
---
### Author:  [@michael-a](https://discuss.elastic.co/u/michael-a)
#### Post date: [December 3, 2025, 2:04pm UTC](https://discuss.elastic.co/t/recommendations-for-llm-nodes-sizing/383813/7 "2025-12-03T14:04:50Z")
Thanks we’ll start small and grow for memory, 16 GB sounds like a good starting point. Are there anything similar for cpu/cores besides the “We recommend that ML nodes have at least 4 Cores and 64GB RAM” from the article? It’s just what we can indicate to our colleagues to let them know what to expect..