llamaperf

DeepSeek V4 Flash 284B (13B active)

on AMD Strix Halo 128GB · 65,536 ctx

Tone: mixedAMD hardware
Sep 7, 2026
Throughput
35.0 t/s gen

Summary

User asks whether anyone has run DeepSeek-V4-Flash-Strix-Halo-GGUF on a Strix Halo. The user notes it fits with ~64K context and is reported at 35 t/s. The user is cautiously optimistic about the result.