Qwen3.5-35B-A3B · IQ1_S
This page aggregates 1 real-world runs of Qwen3.5-35B-A3B (IQ1_S) on NVIDIA RTX 5090 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
227.85 tok/s
Decode
Decode speed
830.74 tok/s
Prefill
Prefill speed
1.34 s
TTFT
Time to first token
30.4 GB
VRAM
VRAM usage
1
Measured runs
1
Independent sources
Other
Source platforms
Today
Last verified
Performance
- vLLM · IQ1_S (current)Decode 227.85 · Prefill 830.74 ·
- llama.cpp · Q6_K Decode 44.2 · Prefill 331.5 ·
- llama.cpp · Q5_K_S Decode 57.5 · Prefill 431.3 ·
- llama.cpp · IQ3_S Decode 95.7 · Prefill 717.8 ·
- llama.cpp · IQ4_XS Decode 72 · Prefill 540 ·
- vLLM · Q4_K_M Decode 69.6 · Prefill 522 ·
Core figures
- Decode (avg)
- 227.85 tok/s
- Prefill (avg)
- 830.74 tok/s
- TTFT (avg)
- 1.34 s
- VRAM (avg)
- 30.4 GB
- MTP acceptance rate
- —
- TTFB
- 1.82 GB
- Power draw
- 454.4 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.5-35B-A3B
- Quantization
- IQ1_S
- Framework
- vLLM
- Version
- v0.31
- 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)
- 30.4 GB
- OS
- Ubuntu 24.04
- Driver
- 570.86
- CUDA
- 12.6
- Power draw
- 454.4 W
Get started
Original model resources
Base model resources may not match this quantization. Use the verified deployment weights above when available.
Sources & evidence
1 measured records in total, each traceable to its original source
v0.31 · Ubuntu 24.04 · CUDA 12.6 · 4096 ctx
227.85 tok/s
Decode
830.74 tok/s
Prefill
1.34 s
TTFT
30.4 GB
VRAM
—
MTP
454.4 W
Power draw
L2 batch seed 2026-10-10: vLLM on hw_id=13 running model_id=15 at IQ1_S