Back to search results
Qwen3.5-0.8B · Q4_K_M
NVIDIA H2097GBvLLM v0.24L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3.5-0.8B (Q4_K_M) on NVIDIA H20 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
121.55 tok/s
Decode
Decode speed
512.9 tok/s
Prefill
Prefill speed
1.06 s
TTFT
Time to first token
3.48 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
16 days ago
Last verified
Performance
- llama.cpp · Q3_K_L Decode 129.02 · Prefill 590.22 ·
- vLLM · Q4_K_M (current)Decode 121.55 · Prefill 512.9 ·
- vLLM · Q8_0 Decode 82.63 · Prefill 440.06 ·
Core figures
- Decode (avg)
- 121.55 tok/s
- Prefill (avg)
- 512.9 tok/s
- TTFT (avg)
- 1.06 s
- VRAM (avg)
- 3.48 GB
- MTP acceptance rate
- —
- TTFB
- 0.97 GB
- Power draw
- 390.9 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.5-0.8B
- Quantization
- Q4_K_M
- Framework
- vLLM
- Version
- v0.24
- Context length
- 8192 tokens
- Batch size
- 32
- GPU layers
- 999
- 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)
- 3.48 GB
- OS
- Windows 11 24H2
- Driver
- 570.86
- CUDA
- 12.6
- Power draw
- 390.9 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.24 · Windows 11 24H2 · CUDA 12.6 · 8192 ctx
121.55 tok/s
Decode
512.9 tok/s
Prefill
1.06 s
TTFT
3.48 GB
VRAM
—
MTP
390.9 W
Power draw
L2 batch seed 2026-10-19: vLLM on hw_id=56 running model_id=51 at Q4_K_M