Arao AI
Back to search results

Llama-3.1-8B-Instruct · IQ4_XS

Start a discussion
NVIDIA RTX PRO 6000 Blackwell 96GB96GBStrata v0.37L2 Cross-framework verified

This page aggregates 1 real-world runs of Llama-3.1-8B-Instruct (IQ4_XS) on NVIDIA RTX PRO 6000 Blackwell 96GB with Strata, contributed by 1 independent source platforms; metrics are averages of published measurements.

387.3 tok/s

Decode

Decode speed

1,749.15 tok/s

Prefill

Prefill speed

1.62 s

TTFT

Time to first token

14.53 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

16 days ago

Last verified

Performance

  1. Ollama · Q3_K_L Decode 262.13 · Prefill 1148.25 ·
  2. Ollama · Q5_K_M Decode 184.14 · Prefill 644.68 ·
  3. Strata · IQ4_XS (current)Decode 387.3 · Prefill 1749.15 ·

Core figures

Decode (avg)
387.3 tok/s
Prefill (avg)
1,749.15 tok/s
TTFT (avg)
1.62 s
VRAM (avg)
14.53 GB
MTP acceptance rate
0.5
TTFB
3.1 GB
Power draw
302.7 W

Configuration

Member-level fields are taken from the most recent run

Model
Llama-3.1-8B-Instruct
Quantization
IQ4_XS
Framework
Strata
Version
v0.37
Context length
4096 tokens
Batch size
512
GPU layers
15
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)
14.53 GB
OS
Ubuntu 22.04
Driver
560.94
CUDA
12.6
Power draw
302.7 W

Sources & evidence

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

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

    v0.37 · Ubuntu 22.04 · CUDA 12.6 · 4096 ctx

    387.3 tok/s

    Decode

    1,749.15 tok/s

    Prefill

    1.62 s

    TTFT

    14.53 GB

    VRAM

    0.5

    MTP

    302.7 W

    Power draw

    L2 batch seed 2026-10-19: Strata on hw_id=68 running model_id=68 at IQ4_XS