# Local LLM model provider — retained record

> Retained reddit record from reddit:homelab.

Canonical: https://highsignal.app/data/records/5ea57edca23fb365

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:homelab)  published 2026-10-22 15:38 ingested 2026-10-23 02:41

[original source ↗](https://www.reddit.com/r/homelab/comments/1xclzlz/local_llm_model_provider/)

## retained text



```
So I've been using ChatGPT with local MCP's for some of my services.  I want to expand my usage of an AI models.  I would like to do both asynchronous tasks (like fetching some data and processing it for me to review later) and synchronous tasks (like "When is the next basketball game for my son").

I have a small homelab PC in which I run many docker containers (the usuals, like Vaultwarden, Mealie, etc.).  But that PC is not beefy enough to run AI workloads on it.  So I'm at a crossroads of whether to run a local model provider (in which I'd need more hardware) or use a cloud provider (like OpenAI, Anthropic, or Grok).   At this point in time, it seems to get any decent performance from a local model you have to be purchasing some expensive hardware.  And that hardware is likely power hungry.

  
So I ask this community:

* For local LLMs: What hardware are you running for local LLMs?  How much did the hardware cost?  What is the electricity cost and is it worth it?
* For Cloud LLMs: Which provider are you using?  What's the cost per month? And are you hitting token limits often and having to spend more money?


```



document fetched 2026-10-23 00:40 · canonical https://reddit.com/r/homelab/comments/1xclzlz/local_llm_model_provider
