Arao AI
Back to search results

mammoth-1 · IQ4_XS

Start a discussion
NVIDIA H2097GBvLLM v0.36L2 Cross-framework verified

This page aggregates 1 real-world runs of mammoth-1 (IQ4_XS) on NVIDIA H20 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

45.95 tok/s

Decode

Decode speed

201.36 tok/s

Prefill

Prefill speed

0.65 s

TTFT

Time to first token

44.06 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

14 days ago

Last verified

Performance

  1. llama.cpp · FP16 Decode 18.55 · Prefill 67.93 ·
  2. vLLM · IQ4_XS (current)Decode 45.95 · Prefill 201.36 ·

Core figures

Decode (avg)
45.95 tok/s
Prefill (avg)
201.36 tok/s
TTFT (avg)
0.65 s
VRAM (avg)
44.06 GB
MTP acceptance rate
—
TTFB
0.37 GB
Power draw
317.7 W

Configuration

Member-level fields are taken from the most recent run

Model
mammoth-1
Quantization
IQ4_XS
Framework
vLLM
Version
v0.36
Context length
8192 tokens
Batch size
16
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 H20
Nominal VRAM
97 GB
Measured VRAM (avg)
44.06 GB
OS
Windows 11 24H2
Driver
570.86
CUDA
12.8
Power draw
317.7 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 2026-10-22

    v0.36 · Windows 11 24H2 · CUDA 12.8 · 8192 ctx

    45.95 tok/s

    Decode

    201.36 tok/s

    Prefill

    0.65 s

    TTFT

    44.06 GB

    VRAM

    —

    MTP

    317.7 W

    Power draw

    L2 batch seed 2026-10-22: vLLM on hw_id=56 running model_id=94 at IQ4_XS