deepseek-v4.1IMAGE

DeepSeek V4.1 Flash on two Strix Halo PCs: 50 tok/s single user, 108 tok/s for 8

Q2 GGUF experts plus ~4-bit trellis dense layers cut bytes per token by 37% with bit-identical output, lifting single-user speed from 45 to 50 tok/s, 70 tok/s on code, and 8-user sharing from 96 to 108 tok/s.

KireiStudios

@KireiStudiosArchived 6 days agoOriginally posted 2026.10.25

DeepSeek V4.1 Flash on two Strix Halo PCs: 50 tok/s single user, 108 tok/s for 8

KireiStudios @KireiStudiosOriginally posted 2026.10.25

Open original on X

Our DeepSeek V4.1 Flash on two small @AIatAMD Strix Halo PCs just got a lot faster: • 50 tok/s for one user (was 45), 70 tok/s on code • 108 tok/s shared across 8 users (was 96) • Same exact output, bit for bit Q2 GGUF experts + our ~4-bit trellis dense layers: 37% fewer bytes per token, same quality.

x.com/kireistudios/status/2114097639771763005Archived 2026.10.25 12:00