← sources

Black Friday coming up, best cost-effective rig for local inference?

source: reddit (reddit:LocalLLaMA)published 2026-11-03 16:35ingested 2026-11-04 02:41

retained text

Got a potato PC that’s 8 years old and running on GTX 1070 TI. Was able to (barely) run some 32B models until I gave up and started using APIs.

What are people’s recommendations for a cost-effective, decent rig for some light to medium local inference like Qwen 3.8 flash next? How fast would it run? What about image generation?

document fetched 2026-11-04 00:41 · canonical https://reddit.com/r/LocalLLaMA/comments/1xmxrfi/black_friday_coming_up_best_costeffective_rig_for