Arao AI
Back to search results

iFlytek Spark-X2.5-4B · Q3_K_S

Start a discussion
NVIDIA RTX PRO 6000 Blackwell 96GB96GBStrata v0.39L2 Cross-framework verified

This page aggregates 1 real-world runs of iFlytek Spark-X2.5-4B (Q3_K_S) on NVIDIA RTX PRO 6000 Blackwell 96GB with Strata, contributed by 1 independent source platforms; metrics are averages of published measurements.

316.72 tok/s

Decode

Decode speed

1,709.01 tok/s

Prefill

Prefill speed

0.17 s

TTFT

Time to first token

7.56 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

12 days ago

Last verified

Performance

  1. llama.cpp · Q6_K Decode 350.22 · Prefill 1535.88 ·
  2. Strata · Q3_K_S (current)Decode 316.72 · Prefill 1709.01 ·

Core figures

Decode (avg)
316.72 tok/s
Prefill (avg)
1,709.01 tok/s
TTFT (avg)
0.17 s
VRAM (avg)
7.56 GB
MTP acceptance rate
0.6
TTFB
2.53 GB
Power draw
322.6 W

Configuration

Member-level fields are taken from the most recent run

Model
iFlytek Spark-X2.5-4B
Quantization
Q3_K_S
Framework
Strata
Version
v0.39
Context length
8192 tokens
Batch size
1024
GPU layers
5
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)
7.56 GB
OS
Windows 11 23H2
Driver
565.57
CUDA
12.4
Power draw
322.6 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 · Windows 11 23H2 · CUDA 12.4 · 8192 ctx

    316.72 tok/s

    Decode

    1,709.01 tok/s

    Prefill

    0.17 s

    TTFT

    7.56 GB

    VRAM

    0.6

    MTP

    322.6 W

    Power draw

    L2 batch seed 2026-10-24: Strata on hw_id=68 running model_id=75 at Q3_K_S