# Best local model for coding? — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/a4c7d141dd5b9adf

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-09-04 16:53 ingested 2026-09-05 02:42

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1w79x68/best_local_model_for_coding/)

## retained text



```
Do you guys use any local models that can run well on a 3090 + 64GB of RAM, specifically for coding?

I really liked MiniMax M3, but I don’t think I can justify continuing to pay for it, so I’m trying to get back into running LLMs locally.

I’ve seen a lot of people recommending Qwen 3.8, but I’m not sure how well it actually competes with some of the closed-source models when it comes to coding. What are you guys using? I’m fine with MiniMax M2.7-level performance if a local model can come close to it.
```



document fetched 2026-09-05 00:42 · canonical https://reddit.com/r/LocalLLaMA/comments/1w79x68/best_local_model_for_coding
