Gemma 4 26B-A4B

UD-Q4_K_M running on Apple M1 Max with macOS 27.0

PERFORMANCE / TOKENS PER SECONDApple Silicon

Prompt processing645.73tok/s

#117 in index

Token generation43.04tok/s

#147 in index

Run fingerprint

Enough context to make the number useful.

ModelGemma 4 26B-A4BMoE architecture

QuantizationUD-Q4_K_MExact submitted variant

GPUApple M1 Max51.84 GB VRAM

GPU architectureApple SiliconMetal backend

Test context512prompt tokens

Model familyMoESigned ToshLLM app · amd-gpu · Metal

ToshLLM versionv0.87.7App 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.