Arao AI
Back to search results

gemma-4-12B-it · Q5_K_S

Start a discussion
NVIDIA H100 80GB HBM380GBllama.cpp v0.27L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-12B-it (Q5_K_S) on NVIDIA H100 80GB HBM3 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

283.07 tok/s

Decode

Decode speed

1,445.04 tok/s

Prefill

Prefill speed

0.32 s

TTFT

Time to first token

24.78 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

16 days ago

Last verified

Performance

  1. llama.cpp · Q5_K_M Decode 303.32 · Prefill 1265.92 ·
  2. llama.cpp · Q5_K_S (current)Decode 283.07 · Prefill 1445.04 ·
  3. vLLM · IQ3_XS Decode 302.76 · Prefill 1418.76 ·
  4. vLLM · IQ4_XS Decode 358.44 · Prefill 1162.82 ·

Core figures

Decode (avg)
283.07 tok/s
Prefill (avg)
1,445.04 tok/s
TTFT (avg)
0.32 s
VRAM (avg)
24.78 GB
MTP acceptance rate
—
TTFB
2.26 GB
Power draw
618 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-12B-it
Quantization
Q5_K_S
Framework
llama.cpp
Version
v0.27
Context length
32768 tokens
Batch size
512
GPU layers
20
Flash Attention
On

Hardware

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

GPU
NVIDIA H100 80GB HBM3
Nominal VRAM
80 GB
Measured VRAM (avg)
24.78 GB
OS
Ubuntu 24.04
Driver
570.86
CUDA
12.8
Power draw
618 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

    v0.27 · Ubuntu 24.04 · CUDA 12.8 · 32768 ctx

    283.07 tok/s

    Decode

    1,445.04 tok/s

    Prefill

    0.32 s

    TTFT

    24.78 GB

    VRAM

    —

    MTP

    618 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=55 running model_id=65 at Q5_K_S