Arao AI
Back to search results

gemma-4-12B-it · IQ3_S

Start a discussion
NVIDIA RTX 509032GBvLLM v0.43L2 Cross-framework verified

This page aggregates 1 real-world runs of gemma-4-12B-it (IQ3_S) on NVIDIA RTX 5090 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

261.31 tok/s

Decode

Decode speed

1,085.12 tok/s

Prefill

Prefill speed

0.38 s

TTFT

Time to first token

18.73 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. vLLM · IQ3_S (current)Decode 261.31 · Prefill 1085.12 ·
  2. vLLM · Q5_K_S Decode 173.67 · Prefill 792.94 ·
  3. Strata · Q3_K_M Decode 340.87 · Prefill 1525.66 ·
  4. llama.cpp · IQ2_XS Decode 253.43 · Prefill 894.02 ·
  5. llama.cpp · IQ3_S Decode 214.27 · Prefill 831.85 ·
  6. vLLM · IQ1_S Decode 298.32 · Prefill 1152.98 ·
  7. vLLM · IQ2_M Decode 288.51 · Prefill 940.03 ·
  8. Ollama · IQ2_XS Decode 246.6 · Prefill 1083.78 ·
  9. vLLM · Q8_0 Decode 117.2 · Prefill 879 ·
  10. vLLM · Q5_K_M Decode 171.1 · Prefill 1283.3 ·
  11. vLLM · Q4_K_M Decode 208.7 · Prefill 1565.3 ·
  12. llama.cpp · Q8_0 Decode 107.5 · Prefill 806.3 ·
  13. llama.cpp · Q5_K_M Decode 157.1 · Prefill 1178.3 ·
  14. llama.cpp · Q4_K_M Decode 191.5 · Prefill 1436.3 ·

Core figures

Decode (avg)
261.31 tok/s
Prefill (avg)
1,085.12 tok/s
TTFT (avg)
0.38 s
VRAM (avg)
18.73 GB
MTP acceptance rate
—
TTFB
2.09 GB
Power draw
370.7 W

Configuration

Member-level fields are taken from the most recent run

Model
gemma-4-12B-it
Quantization
IQ3_S
Framework
vLLM
Version
v0.43
Context length
4096 tokens
Batch size
8
GPU layers
999
Flash Attention
On

Hardware

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

GPU
NVIDIA RTX 5090
Nominal VRAM
32 GB
Measured VRAM (avg)
18.73 GB
OS
Windows 11 24H2
Driver
565.57
CUDA
12.4
Power draw
370.7 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-18

    v0.43 · Windows 11 24H2 · CUDA 12.4 · 4096 ctx

    261.31 tok/s

    Decode

    1,085.12 tok/s

    Prefill

    0.38 s

    TTFT

    18.73 GB

    VRAM

    —

    MTP

    370.7 W

    Power draw

    L2 batch seed 2026-10-18: vLLM on hw_id=13 running model_id=65 at IQ3_S