llamaperf

Gemma 4 26B (4B active)

on AMD Strix Halo 128GB · Gufo · 2,038 ctx

AMD hardware
Oct 10, 2026
Throughput
53.2 t/s gen · 3991.2 tokens/s prompt processing (prefill)Prompt processing measures input prefill speed; it does not tell you the time to first token.
Quant
UD-Q4_K_XL (GGUF)
MTP (Multi-Token Prediction)
off
System RAM
128 GB

Summary

User reports Gemma 4 26B-A4B at 53.2 t/s generation and 3991.2 t/s prefill on a 2,038-token prompt on an AMD Strix Halo 128GB. Setup is Gufo at UD-Q4_K_XL, one request, speculative decoding off. franzmoca's depth-zero row of 2026-10-07 from the Gemma 4 Gufo fork. A source-reported maintainer measurement, not independently reproduced.