# Best local for RTX 4090 + 3090 Ti + 128GB DDR4? — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/7df6984a3293d34e

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-10-29 09:04 ingested 2026-10-30 02:40

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1xieult/best_local_for_rtx_4090_3090_ti_128gb_ddr4/)

## retained text



```
&#x200B;

Can I run DeepSeek V4.1 Flash 1017?

My setup:

\- RTX 4090 24GB + RTX 3090 Ti 24GB

\- 128GB DDR4 RAM

\- Intel i7-13700KF

What's the smartest and fastest open-weight LLM I can realistically run , especially for reasoning, coding, and medical knowledge?

Can I run DeepSeek V4.1 Flash 1017 using Strata, SSD offloading, or extreme quantization? What tokens/sec should I expect?

Looking for real-world experience and recommendations for the best model, quantization, and inference engine.
```



document fetched 2026-10-30 00:45 · canonical https://reddit.com/r/LocalLLaMA/comments/1xieult/best_local_for_rtx_4090_3090_ti_128gb_ddr4
