DeepSeek V4.1 Flash Runs at 20 tok/s on a Single RTX 3090

Developer casperhansen ran DeepSeek V4.1 Flash 1017 locally on a single RTX 3090 with 64GB RAM at 20 tokens/s with zero quality loss, aided by the official open-source inference stack; adding another 64GB of memory can double the speed.

2026-11-12 ~ 2026-11-12 · 2 related posts