Arao AI
Back to search results

Qwen3-8B · Q2_K

Start a discussion
NVIDIA RTX PRO 6000 Blackwell 96GB96GBllama.cpp v0.27L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3-8B (Q2_K) on NVIDIA RTX PRO 6000 Blackwell 96GB with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

305.21 tok/s

Decode

Decode speed

1,181.07 tok/s

Prefill

Prefill speed

2.29 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

17 days ago

Last verified

Performance

  1. llama.cpp · Q2_K (current)Decode 305.21 · Prefill 1181.07 ·
  2. Strata · IQ3_S Decode 374.23 · Prefill 1951.46 ·
  3. Strata · Q3_K_L Decode 317.22 · Prefill 1349.4 ·

Core figures

Decode (avg)
305.21 tok/s
Prefill (avg)
1,181.07 tok/s
TTFT (avg)
2.29 s
VRAM (avg)
11.34 GB
MTP acceptance rate
—
TTFB
2.44 GB
Power draw
377 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-8B
Quantization
Q2_K
Framework
llama.cpp
Version
v0.27
Context length
4096 tokens
Batch size
512
GPU layers
28
Flash Attention
On

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
Ubuntu 22.04
Driver
560.94
CUDA
12.4
Power draw
377 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 22.04 · CUDA 12.4 · 4096 ctx

    305.21 tok/s

    Decode

    1,181.07 tok/s

    Prefill

    2.29 s

    TTFT

    11.34 GB

    VRAM

    —

    MTP

    377 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=68 running model_id=53 at Q2_K