Back to search results
Llama-3-8B-Instruct · Q2_K
NVIDIA H2097GBvLLM v0.27L2 Cross-framework verified
This page aggregates 1 real-world runs of Llama-3-8B-Instruct (Q2_K) on NVIDIA H20 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
77.65 tok/s
Decode
Decode speed
286.25 tok/s
Prefill
Prefill speed
0.36 s
TTFT
Time to first token
11.34 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
17 days ago
Last verified
Performance
- vLLM · Q2_K (current)Decode 77.65 · Prefill 286.25 ·
- Ollama · Q5_K_M Decode 41.54 · Prefill 137.1 ·
- Ollama · Q8_0 Decode 36.49 · Prefill 129.53 ·
Core figures
- Decode (avg)
- 77.65 tok/s
- Prefill (avg)
- 286.25 tok/s
- TTFT (avg)
- 0.36 s
- VRAM (avg)
- 11.34 GB
- MTP acceptance rate
- —
- TTFB
- 0.62 GB
- Power draw
- 312.8 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Llama-3-8B-Instruct
- Quantization
- Q2_K
- Framework
- vLLM
- Version
- v0.27
- Context length
- 2048 tokens
- Batch size
- 16
- 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)
- 11.34 GB
- OS
- Ubuntu 24.04
- Driver
- 570.86
- CUDA
- 12.4
- Power draw
- 312.8 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.27 · Ubuntu 24.04 · CUDA 12.4 · 2048 ctx
77.65 tok/s
Decode
286.25 tok/s
Prefill
0.36 s
TTFT
11.34 GB
VRAM
—
MTP
312.8 W
Power draw
L2 batch seed 2026-10-19: vLLM on hw_id=56 running model_id=69 at Q2_K