gemma-4-26B-A4B-it on Intel Arc Pro B65 32GB

docker run --rm -d --name lsv-gem82-u-tp1-ctx131k --net=host --shm-size=32g --device /dev/dri:/dev/dri -v /dev/dri:/dev/dri -v /models/weights:/llm/models:ro -v /models/bench/lmx-ctx-20261019/session-a/gem82-u-tp1-ctx131k/session:/bench -e ZE_AFFINITY_MASK=0 -e VLLM_XPU_ENABLE_XPU_GRAPH=1 -e DISABLE_ESIMD_GDN_OUTPROJ=0 -e DISABLE_ESIMD_INT4=0 -e DISABLE_ESIMD_INT4_GEMM=0 -e VLLM_OFFLOAD_WEIGHTS_BEFORE_QUANT=1 -e VLLM_ALLOW_LONG_MAX_MODEL_LEN=1 -e VLLM_WORKER_MULTIPROC_METHOD=spawn -e CCL_ZE_IPC_EXCHANGE=sockets --entrypoint vllm vllm/vllm-openai-xpu:v0.32.0@sha256:20509fabe50dda62c09c0786d1a6effa299d3dbdddcc49331b93e65144652a8c serve --model /llm/models/gemma-4-26B-A4B-it --host 0.0.0.0 --served-model-name gemma-4 --tensor-parallel-size 1 --pipeline-parallel-size 1 --data-parallel-size 1 --gpu-memory-utilization 0.97 --max-model-len 5120 --max-num-seqs 16 --max-num-batched-tokens 8192 --block-size 64 --kv-cache-dtype auto --mamba-ssm-cache-dtype auto --dtype bfloat16 --quantization fp8 --enable-prefix-caching --trust-remote-code --limit-mm-per-prompt \{\"image\":0\,\"video\":0\} --cpu-offload-gb 0 --port 8000

TTFT is measured after the benchmark warms the server's prompt cache, so it mostly excludes prefill.