← sources

Best local model for coding?

source: reddit (reddit:LocalLLaMA)published 2026-09-04 16:53ingested 2026-09-05 02:42

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