Gemma 4 26B-A4B · vLLM · NVFP4 · conc 2

status: done

Configuration

Modelgoogle/gemma-4-26B-A4B-it
CompanyGoogle
FamilyGemma
Parameters26B / 4B (MoE)
EnginevLLM
Quant / precisionNVFP4
Why this quantNVIDIA's own NVFP4 build (TensorRT-Model-Optimizer / modelopt) — Blackwell-native 4-bit. The autoregressive reference line (no drafter) for the decode-vs-concurrency crossover figure.
Downloadnvidia/Gemma-4-26B-A4B-NVFP4
Context window65536
Input modalitiestext, image (served text-only here)

Measured results

Prefill tok/s88.33
Decode tok/s61.97
Peak memory (GB)109.11 (system MemAvailable delta (10s sampling) — vLLM static KV reservation (util 0.85))
Completed2026-07-03 15:15 +0800

Full run command

VLLM_IMAGE=vllm/vllm-openai:nightly-aarch64 scripts/bench-vllm-serving.sh nvidia/Gemma-4-26B-A4B-NVFP4 65536 2 400 400 256
# 111/400 prompts (hit the time cap), 0 errors, 408.3s. ready after 292s. TTFT median 123.7 ms, TPOT median 31.0 ms, req thr 0.272/s.

gemma-4-26b-a4b Google Gemma NVFP4 16-40B conc-2

Notes

Decode 61.97 tok/s aggregate at concurrency 2. Autoregressive NVFP4 reference point, conc 2, for the decode-vs-concurrency crossover figure.

  • Image (pinned): vllm/vllm-openai:nightly-aarch64 @ sha256:e414712fdc04… — the SINGLE image for all 24 cells of this figure. Ready after 292 s.
  • Workload: ShareGPT V3, concurrency 2. 111/400 completed, 0 errors before the 400 s time cap.
  • Throughput: decode 61.97 tok/s aggregate, prefill 88.33 tok/s. TTFT median 123.7 ms, TPOT median 31.0 ms, req throughput 0.272/s.
  • Memory: 109.11 GB = vLLM --gpu-memory-utilization 0.85 reservation, not the model footprint.