Back to search results
Qwen3-Coder-Next · Q3_K_S
2x AMD Strix HalovLLM v0.48L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3-Coder-Next (Q3_K_S) on 2x AMD Strix Halo with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
149.19 tok/s
Decode
Decode speed
617.1 tok/s
Prefill
Prefill speed
1.16 s
TTFT
Time to first token
1.05 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
2 days ago
Last verified
Performance
- vLLM · Q3_K_S (current)Decode 149.19 · Prefill 617.1 ·
- vLLM · Q8_0 Decode 87.69 · Prefill 408.22 ·
- Strata · Q3_K_M Decode 208.39 · Prefill 864.96 ·
Core figures
- Decode (avg)
- 149.19 tok/s
- Prefill (avg)
- 617.1 tok/s
- TTFT (avg)
- 1.16 s
- VRAM (avg)
- 1.05 GB
- MTP acceptance rate
- —
- TTFB
- 1.19 GB
- Power draw
- 98.4 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3-Coder-Next
- Quantization
- Q3_K_S
- Framework
- vLLM
- Version
- v0.48
- Context length
- 8192 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
- 2x AMD Strix Halo
- Measured VRAM (avg)
- 1.05 GB
- OS
- Ubuntu 22.04
- Driver
- 560.94
- CUDA
- N/A
- Power draw
- 98.4 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.48 · Ubuntu 22.04 · CUDA N/A · 8192 ctx
149.19 tok/s
Decode
617.1 tok/s
Prefill
1.16 s
TTFT
1.05 GB
VRAM
—
MTP
98.4 W
Power draw
L2 batch seed 2026-10-16: vLLM on hw_id=88 running model_id=58 at Q3_K_S