Gemma4 26B-A4B

Q8_K running on AMD Radeon RX 580 with macOS 15.7.3 Sequoia

PERFORMANCE / TOKENS PER SECONDGCN / Vega

Prompt processing46.59tok/s

#342 in index

Token generation5.51tok/s

#339 in index

Run fingerprint

Enough context to make the number useful.

ModelGemma4 26B-A4BMoE architecture

QuantizationQ8_KExact submitted variant

GPUAMD Radeon RX 5808 GB VRAM

GPU architectureGCN / VegaMetal backend

Test context512prompt tokens

Model familyMoESigned ToshLLM app · ncmoe 40 · amd-gpu · Metal

ToshLLM versionv0.87.0App 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.