Arao AI
Back to search results

Qwen3.6-27B · Q8_0

Start a discussion
NVIDIA RTX PRO 6000 Blackwell 96GB96GBllama.cpp v0.48L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.6-27B (Q8_0) on NVIDIA RTX PRO 6000 Blackwell 96GB with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

105.75 tok/s

Decode

Decode speed

445.38 tok/s

Prefill

Prefill speed

1.17 s

TTFT

Time to first token

67.3 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 · IQ1_S Decode 267.34 · Prefill 934.02 ·
  2. llama.cpp · Q8_0 (current)Decode 105.75 · Prefill 445.38 ·
  3. vLLM · IQ3_S Decode 220.08 · Prefill 989.75 ·

Core figures

Decode (avg)
105.75 tok/s
Prefill (avg)
445.38 tok/s
TTFT (avg)
1.17 s
VRAM (avg)
67.3 GB
MTP acceptance rate
—
TTFB
0.85 GB
Power draw
375.7 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.6-27B
Quantization
Q8_0
Framework
llama.cpp
Version
v0.48
Context length
2048 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 RTX PRO 6000 Blackwell 96GB
Nominal VRAM
96 GB
Measured VRAM (avg)
67.3 GB
OS
Ubuntu 22.04
Driver
565.57
CUDA
12.4
Power draw
375.7 W

Get started

Original model resources

Base model resources may not match this quantization. Use the verified deployment weights above when available.

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-19

    v0.48 · Ubuntu 22.04 · CUDA 12.4 · 2048 ctx

    105.75 tok/s

    Decode

    445.38 tok/s

    Prefill

    1.17 s

    TTFT

    67.3 GB

    VRAM

    —

    MTP

    375.7 W

    Power draw

    L2 batch seed 2026-10-19: llama.cpp on hw_id=68 running model_id=6 at Q8_0