Back to search results
Qwen3-8B · IQ4_XS
NVIDIA RTX 509032GBvLLM v0.42L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3-8B (IQ4_XS) on NVIDIA RTX 5090 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
252.51 tok/s
Decode
Decode speed
829.1 tok/s
Prefill
Prefill speed
1.53 s
TTFT
Time to first token
14.53 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
Today
Last verified
Performance
- llama.cpp · IQ4_XS Decode 253.29 · Prefill 909.82 ·
- llama.cpp · Q6_K Decode 179.01 · Prefill 827.24 ·
- vLLM · IQ3_S Decode 284.83 · Prefill 1483.72 ·
- vLLM · IQ4_XS (current)Decode 252.51 · Prefill 829.1 ·
- Ollama · Q5_K_M Decode 172.9 · Prefill 940.17 ·
- llama.cpp · Q8_0 Decode 162.5 · Prefill 1218.8 ·
- llama.cpp · Q5_K_M Decode 236.9 · Prefill 1776.8 ·
- llama.cpp · Q4_K_M Decode 285.3 · Prefill 2139.8 ·
Core figures
- Decode (avg)
- 252.51 tok/s
- Prefill (avg)
- 829.1 tok/s
- TTFT (avg)
- 1.53 s
- VRAM (avg)
- 14.53 GB
- MTP acceptance rate
- —
- TTFB
- 2.02 GB
- Power draw
- 435.9 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3-8B
- Quantization
- IQ4_XS
- Framework
- vLLM
- Version
- v0.42
- Context length
- 4096 tokens
- Batch size
- 8
- GPU layers
- 999
- Flash Attention
- Off
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)
- 14.53 GB
- OS
- Ubuntu 24.04
- Driver
- 555.85
- CUDA
- 12.8
- Power draw
- 435.9 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.42 · Ubuntu 24.04 · CUDA 12.8 · 4096 ctx
252.51 tok/s
Decode
829.1 tok/s
Prefill
1.53 s
TTFT
14.53 GB
VRAM
—
MTP
435.9 W
Power draw
L2 batch seed 2026-10-10: vLLM on hw_id=13 running model_id=53 at IQ4_XS