Qwen3.5-0.8B · Q5_K_M
This page aggregates 1 real-world runs of Qwen3.5-0.8B (Q5_K_M) on NVIDIA RTX Spark with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
99.17 tok/s
Decode
Decode speed
509.83 tok/s
Prefill
Prefill speed
1.79 s
TTFT
Time to first token
3.66 GB
VRAM
VRAM usage
1
Measured runs
1
Independent sources
Other
Source platforms
17 days ago
Last verified
Performance
- llama.cpp · Q6_K Decode 94.74 · Prefill 304.72 ·
- llama.cpp · Q8_0 Decode 76.23 · Prefill 286.92 ·
- Ollama · FP16 Decode 57.1 · Prefill 306 ·
- Ollama · Q8_0 Decode 79.06 · Prefill 361.81 ·
- llama.cpp · Q4_K_M Decode 129.1 · Prefill 701.36 ·
- llama.cpp · Q5_K_M (current)Decode 99.17 · Prefill 509.83 ·
- Ollama · Q4_K_M Decode 109.43 · Prefill 483.78 ·
Core figures
- Decode (avg)
- 99.17 tok/s
- Prefill (avg)
- 509.83 tok/s
- TTFT (avg)
- 1.79 s
- VRAM (avg)
- 3.66 GB
- MTP acceptance rate
- —
- TTFB
- 0.79 GB
- Power draw
- 73 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.5-0.8B
- Quantization
- Q5_K_M
- Framework
- llama.cpp
- Version
- v0.47
- Context length
- 4096 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 RTX Spark
- Nominal VRAM
- 128 GB
- Measured VRAM (avg)
- 3.66 GB
- OS
- Ubuntu 24.04
- Driver
- NVIDIA 590.10
- CUDA
- 13.0
- Power draw
- 73 W
- Reference price
- $2,399 · updated 2026-10-19
Prices are not live; check the store page
Sources & evidence
1 measured records in total, each traceable to its original source
v0.47 · Ubuntu 24.04 · CUDA 13.0 · 4096 ctx
99.17 tok/s
Decode
509.83 tok/s
Prefill
1.79 s
TTFT
3.66 GB
VRAM
—
MTP
73 W
Power draw
L2 batch seed 2026-10-19: llama.cpp on hw_id=110 running model_id=51 at Q5_K_M