Gemma 4 26B-A4B

UD-Q4_K_M running on AMD Radeon Pro 5700 XT with macOS 15.6.1 Sequoia

PERFORMANCE / TOKENS PER SECONDRDNA 1

Prompt processing424.08tok/s

#113 in index

Token generation21.9tok/s

#141 in index

Run fingerprint

Enough context to make the number useful.

ModelGemma 4 26B-A4BMoE architecture

QuantizationUD-Q4_K_MExact submitted variant

GPUAMD Radeon Pro 5700 XT15.98 GB VRAM

GPU architectureRDNA 1Metal backend

Test context512prompt tokens

Model familyMoESigned ToshLLM app · ncmoe 16 · amd-gpu · Metal

ToshLLM versionv0.83.17App version used to capture and sign this run

Prompt processing and generation are reported separately so memory bandwidth and decode behavior stay visible.

Quantization, GPU family, VRAM and context travel with the result instead of being hidden behind one score.

This result remains explicitly labeled as a community submission.