Arao AI
Back to search results

Ministral-3-8B-Instruct-2512 · IQ3_XS

Start a discussion
NVIDIA RTX 4080 Super16GBvLLM v0.39L2 Cross-framework verified

This page aggregates 1 real-world runs of Ministral-3-8B-Instruct-2512 (IQ3_XS) on NVIDIA RTX 4080 Super with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

194.23 tok/s

Decode

Decode speed

865.43 tok/s

Prefill

Prefill speed

0.52 s

TTFT

Time to first token

12.52 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

12 days ago

Last verified

Performance

  1. vLLM · IQ3_XS (current)Decode 194.23 · Prefill 865.43 ·
  2. Ollama · Q3_K_S Decode 155.17 · Prefill 518.65 ·

Core figures

Decode (avg)
194.23 tok/s
Prefill (avg)
865.43 tok/s
TTFT (avg)
0.52 s
VRAM (avg)
12.52 GB
MTP acceptance rate
—
TTFB
1.55 GB
Power draw
299 W

Configuration

Member-level fields are taken from the most recent run

Model
Ministral-3-8B-Instruct-2512
Quantization
IQ3_XS
Framework
vLLM
Version
v0.39
Context length
8192 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 RTX 4080 Super
Nominal VRAM
16 GB
Measured VRAM (avg)
12.52 GB
OS
Ubuntu 22.04
Driver
572.83
CUDA
12.8
Power draw
299 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-24

    v0.39 · Ubuntu 22.04 · CUDA 12.8 · 8192 ctx

    194.23 tok/s

    Decode

    865.43 tok/s

    Prefill

    0.52 s

    TTFT

    12.52 GB

    VRAM

    —

    MTP

    299 W

    Power draw

    L2 batch seed 2026-10-24: vLLM on hw_id=20 running model_id=74 at IQ3_XS