Arao AI
Back to search results

Ministral-3-8B-Instruct-2512 · Q2_K

Start a discussion
NVIDIA RTX PRO 6000 Blackwell 96GB96GBvLLM v0.35L2 Cross-framework verified

This page aggregates 1 real-world runs of Ministral-3-8B-Instruct-2512 (Q2_K) on NVIDIA RTX PRO 6000 Blackwell 96GB with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

326.52 tok/s

Decode

Decode speed

1,378.65 tok/s

Prefill

Prefill speed

2.24 s

TTFT

Time to first token

11.34 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

11 days ago

Last verified

Performance

  1. vLLM · Q2_K (current)Decode 326.52 · Prefill 1378.65 ·
  2. Strata · Q5_K_M Decode 252.2 · Prefill 1116.29 ·

Core figures

Decode (avg)
326.52 tok/s
Prefill (avg)
1,378.65 tok/s
TTFT (avg)
2.24 s
VRAM (avg)
11.34 GB
MTP acceptance rate
—
TTFB
2.61 GB
Power draw
329.4 W

Configuration

Member-level fields are taken from the most recent run

Model
Ministral-3-8B-Instruct-2512
Quantization
Q2_K
Framework
vLLM
Version
v0.35
Context length
65536 tokens
Batch size
32
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 PRO 6000 Blackwell 96GB
Nominal VRAM
96 GB
Measured VRAM (avg)
11.34 GB
OS
Windows 11 23H2
Driver
572.83
CUDA
12.8
Power draw
329.4 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

    v0.35 · Windows 11 23H2 · CUDA 12.8 · 65536 ctx

    326.52 tok/s

    Decode

    1,378.65 tok/s

    Prefill

    2.24 s

    TTFT

    11.34 GB

    VRAM

    —

    MTP

    329.4 W

    Power draw

    L2 batch seed 2026-10-24: vLLM on hw_id=68 running model_id=74 at Q2_K