Back to search results
Qwen3.5-4B · IQ2_XS
NVIDIA RTX 509032GBOllama v0.32L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3.5-4B (IQ2_XS) on NVIDIA RTX 5090 with Ollama, contributed by 1 independent source platforms; metrics are averages of published measurements.
333.52 tok/s
Decode
Decode speed
1,258.8 tok/s
Prefill
Prefill speed
0.89 s
TTFT
Time to first token
6.13 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
2 days ago
Last verified
Performance
- llama.cpp · IQ2_M Decode 354.14 · Prefill 1511.04 ·
- llama.cpp · Q4_K_S Decode 353.61 · Prefill 1327.69 ·
- Ollama · IQ2_M Decode 347.38 · Prefill 1126.85 ·
- Ollama · IQ2_XS (current)Decode 333.52 · Prefill 1258.8 ·
- llama.cpp · Q3_K_L Decode 338.21 · Prefill 1409.78 ·
- vLLM · Q3_K_M Decode 356.1 · Prefill 1633.97 ·
- vLLM · Q3_K_S Decode 324.89 · Prefill 1253.12 ·
- Strata · Q4_K_S Decode 353.67 · Prefill 1676.75 ·
Core figures
- Decode (avg)
- 333.52 tok/s
- Prefill (avg)
- 1,258.8 tok/s
- TTFT (avg)
- 0.89 s
- VRAM (avg)
- 6.13 GB
- MTP acceptance rate
- —
- TTFB
- 2.67 GB
- Power draw
- 384.3 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.5-4B
- Quantization
- IQ2_XS
- Framework
- Ollama
- Version
- v0.32
- Context length
- 8192 tokens
- Batch size
- 512
- GPU layers
- 20
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- NVIDIA RTX 5090
- Nominal VRAM
- 32 GB
- Measured VRAM (avg)
- 6.13 GB
- OS
- macOS 15.0
- Driver
- 560.94
- CUDA
- 12.8
- Power draw
- 384.3 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.32 · macOS 15.0 · CUDA 12.8 · 8192 ctx
333.52 tok/s
Decode
1,258.8 tok/s
Prefill
0.89 s
TTFT
6.13 GB
VRAM
—
MTP
384.3 W
Power draw
L2 batch seed 2026-10-16: Ollama on hw_id=13 running model_id=48 at IQ2_XS