Arao AI
Back to search results

Hy-Memory-30B-A3B · IQ3_XS

Start a discussion
NVIDIA H2097GBStrata v0.42L2 Cross-framework verified

This page aggregates 1 real-world runs of Hy-Memory-30B-A3B (IQ3_XS) on NVIDIA H20 with Strata, contributed by 1 independent source platforms; metrics are averages of published measurements.

139.08 tok/s

Decode

Decode speed

595.25 tok/s

Prefill

Prefill speed

1.43 s

TTFT

Time to first token

41.16 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 · Q6_K Decode 57.59 · Prefill 253.97 ·
  2. Strata · IQ1_S Decode 178.17 · Prefill 571.19 ·
  3. Strata · IQ3_XS (current)Decode 139.08 · Prefill 595.25 ·

Core figures

Decode (avg)
139.08 tok/s
Prefill (avg)
595.25 tok/s
TTFT (avg)
1.43 s
VRAM (avg)
41.16 GB
MTP acceptance rate
0.57
TTFB
1.11 GB
Power draw
380.9 W

Configuration

Member-level fields are taken from the most recent run

Model
Hy-Memory-30B-A3B
Quantization
IQ3_XS
Framework
Strata
Version
v0.42
Context length
4096 tokens
Batch size
1024
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 H20
Nominal VRAM
97 GB
Measured VRAM (avg)
41.16 GB
OS
Windows 11 23H2
Driver
560.94
CUDA
12.6
Power draw
380.9 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

    v0.42 · Windows 11 23H2 · CUDA 12.6 · 4096 ctx

    139.08 tok/s

    Decode

    595.25 tok/s

    Prefill

    1.43 s

    TTFT

    41.16 GB

    VRAM

    0.57

    MTP

    380.9 W

    Power draw

    L2 batch seed 2026-10-19: Strata on hw_id=56 running model_id=87 at IQ3_XS