← sources

Budget local setup for agentic use

source: reddit (reddit:LocalLLaMA)published 2026-10-26 13:55ingested 2026-10-27 02:39

retained text

I’m planning to seup a local LLM build and need advice if it is good to start or recommend if you think it's not good.

My current option is a Ryzen 9900X + RTX 3090 + 64GB RAM. Or I could get a cheaper CPU, such as the 9700X or lower, and put the savings for more RAM later.

A few questions:

1. Is the 9900X worth the extra cost for this workload, or cheaper CPU would be sufficient?
2. Is it worth paying more for a motherboard that supports two gpu slots, so I can add another GPU later? Or should I prioritize more RAM?
3. main use is agentic coding, with a long context window. Could this setup run models close to GPT-6 Luna low or Haiku 5.5 low?

I’m trying to avoid spending money on a build that technically runs the model but is too slow or unreliable for practical agentic use.

document fetched 2026-10-27 00:43 · canonical https://reddit.com/r/LocalLLaMA/comments/1xfxhpf/budget_local_setup_for_agentic_use