Arao AI
Back to search results

gemma-4-31B-it · Q4_K_S

Start a discussion
NVIDIA H2097GBvLLM v0.49L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-31B-it (Q4_K_S) on NVIDIA H20 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

32.57 tok/s

Decode

Decode speed

141.17 tok/s

Prefill

Prefill speed

2.26 s

TTFT

Time to first token

52.88 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

16 days ago

Last verified

Performance

  1. vLLM · IQ3_XS Decode 39.95 · Prefill 179.27 ·
  2. vLLM · Q4_K_S (current)Decode 32.57 · Prefill 141.17 ·
  3. Ollama · Q8_0 Decode 18.96 · Prefill 65.76 ·

Core figures

Decode (avg)
32.57 tok/s
Prefill (avg)
141.17 tok/s
TTFT (avg)
2.26 s
VRAM (avg)
52.88 GB
MTP acceptance rate
—
TTFB
0.26 GB
Power draw
339.3 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-31B-it
Quantization
Q4_K_S
Framework
vLLM
Version
v0.49
Context length
2048 tokens
Batch size
16
GPU layers
999
Flash Attention
Off

Hardware

Nominal and measured figures are shown side by side; whether it runs is the reader's call

GPU
NVIDIA H20
Nominal VRAM
97 GB
Measured VRAM (avg)
52.88 GB
OS
Ubuntu 22.04
Driver
560.94
CUDA
12.8
Power draw
339.3 W

Sources & evidence

1 measured records in total, each traceable to its original source

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-19

    v0.49 · Ubuntu 22.04 · CUDA 12.8 · 2048 ctx

    32.57 tok/s

    Decode

    141.17 tok/s

    Prefill

    2.26 s

    TTFT

    52.88 GB

    VRAM

    —

    MTP

    339.3 W

    Power draw

    L2 batch seed 2026-10-19: vLLM on hw_id=56 running model_id=63 at Q4_K_S