Arao AI
Back to search results

Hy-Memory-30B-A3B · IQ2_XS

Start a discussion
NVIDIA H100 80GB HBM380GBvLLM v0.24L2 Cross-framework verified

This page aggregates 1 real-world runs of Hy-Memory-30B-A3B (IQ2_XS) on NVIDIA H100 80GB HBM3 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

346.16 tok/s

Decode

Decode speed

1,601.93 tok/s

Prefill

Prefill speed

2.02 s

TTFT

Time to first token

32.34 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 · Q6_K Decode 289.84 · Prefill 1568.48 ·
  2. vLLM · IQ2_XS (current)Decode 346.16 · Prefill 1601.93 ·
  3. vLLM · Q5_K_M Decode 299.81 · Prefill 1193.35 ·

Core figures

Decode (avg)
346.16 tok/s
Prefill (avg)
1,601.93 tok/s
TTFT (avg)
2.02 s
VRAM (avg)
32.34 GB
MTP acceptance rate
—
TTFB
2.77 GB
Power draw
688.5 W

Configuration

Member-level fields are taken from the most recent run

Model
Hy-Memory-30B-A3B
Quantization
IQ2_XS
Framework
vLLM
Version
v0.24
Context length
4096 tokens
Batch size
32
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 H100 80GB HBM3
Nominal VRAM
80 GB
Measured VRAM (avg)
32.34 GB
OS
Ubuntu 24.04
Driver
560.94
CUDA
12.8
Power draw
688.5 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

    v0.24 · Ubuntu 24.04 · CUDA 12.8 · 4096 ctx

    346.16 tok/s

    Decode

    1,601.93 tok/s

    Prefill

    2.02 s

    TTFT

    32.34 GB

    VRAM

    —

    MTP

    688.5 W

    Power draw

    L2 batch seed 2026-10-19: vLLM on hw_id=55 running model_id=87 at IQ2_XS