Back to search results
Qwen3.5-0.8B · Q3_K_L
NVIDIA H2097GBllama.cpp v0.24L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3.5-0.8B (Q3_K_L) on NVIDIA H20 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
129.02 tok/s
Decode
Decode speed
590.22 tok/s
Prefill
Prefill speed
0.86 s
TTFT
Time to first token
3.31 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
17 days ago
Last verified
Performance
- llama.cpp · Q3_K_L (current)Decode 129.02 · Prefill 590.22 ·
- vLLM · Q4_K_M Decode 121.55 · Prefill 512.9 ·
- vLLM · Q8_0 Decode 82.63 · Prefill 440.06 ·
Core figures
- Decode (avg)
- 129.02 tok/s
- Prefill (avg)
- 590.22 tok/s
- TTFT (avg)
- 0.86 s
- VRAM (avg)
- 3.31 GB
- MTP acceptance rate
- —
- TTFB
- 1.03 GB
- Power draw
- 324.8 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.5-0.8B
- Quantization
- Q3_K_L
- Framework
- llama.cpp
- Version
- v0.24
- Context length
- 16384 tokens
- Batch size
- 512
- GPU layers
- 33
- 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)
- 3.31 GB
- OS
- Ubuntu 22.04
- Driver
- 570.86
- CUDA
- 12.6
- Power draw
- 324.8 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.24 · Ubuntu 22.04 · CUDA 12.6 · 16384 ctx
129.02 tok/s
Decode
590.22 tok/s
Prefill
0.86 s
TTFT
3.31 GB
VRAM
—
MTP
324.8 W
Power draw
L2 batch seed 2026-10-19: llama.cpp on hw_id=56 running model_id=51 at Q3_K_L