Back to search results
Qwen3.5-4B · Q5_K_S
NVIDIA RTX 507012GBOllama v0.32L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3.5-4B (Q5_K_S) on NVIDIA RTX 5070 with Ollama, contributed by 1 independent source platforms; metrics are averages of published measurements.
83.31 tok/s
Decode
Decode speed
418.84 tok/s
Prefill
Prefill speed
1.41 s
TTFT
Time to first token
9.66 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
Today
Last verified
Performance
- Ollama · Q3_K_S Decode 115.6 · Prefill 548.24 ·
- Ollama · Q5_K_S (current)Decode 83.31 · Prefill 418.84 ·
- vLLM · Q3_K_M Decode 112.72 · Prefill 436.5 ·
- vLLM · Q4_K_M Decode 84.51 · Prefill 408.98 ·
- Strata · IQ3_S Decode 154.26 · Prefill 563.04 ·
- llama.cpp · Q5_K_M Decode 174.7 · Prefill 1310.3 ·
Core figures
- Decode (avg)
- 83.31 tok/s
- Prefill (avg)
- 418.84 tok/s
- TTFT (avg)
- 1.41 s
- VRAM (avg)
- 9.66 GB
- MTP acceptance rate
- —
- TTFB
- 0.67 GB
- Power draw
- 244.5 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.5-4B
- Quantization
- Q5_K_S
- Framework
- Ollama
- Version
- v0.32
- Context length
- 16384 tokens
- Batch size
- 256
- GPU layers
- 99
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- NVIDIA RTX 5070
- Nominal VRAM
- 12 GB
- Measured VRAM (avg)
- 9.66 GB
- OS
- Windows 11 24H2
- Driver
- 572.83
- CUDA
- 13.0
- Power draw
- 244.5 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.32 · Windows 11 24H2 · CUDA 13.0 · 16384 ctx
83.31 tok/s
Decode
418.84 tok/s
Prefill
1.41 s
TTFT
9.66 GB
VRAM
—
MTP
244.5 W
Power draw
L2 batch seed 2026-10-18: Ollama on hw_id=18 running model_id=48 at Q5_K_S