Back to search results
Qwen2.5-Omni-3B · IQ2_XS
2x AMD Strix Halollama.cpp v0.37L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen2.5-Omni-3B (IQ2_XS) on 2x AMD Strix Halo with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
273.09 tok/s
Decode
Decode speed
1,000.97 tok/s
Prefill
Prefill speed
1.79 s
TTFT
Time to first token
5.12 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_XS (current)Decode 273.09 · Prefill 1000.97 ·
- llama.cpp · IQ3_S Decode 239.07 · Prefill 794.3 ·
- vLLM · Q5_K_M Decode 183.47 · Prefill 864.96 ·
- vLLM · Q6_K Decode 171.22 · Prefill 756.33 ·
Core figures
- Decode (avg)
- 273.09 tok/s
- Prefill (avg)
- 1,000.97 tok/s
- TTFT (avg)
- 1.79 s
- VRAM (avg)
- 5.12 GB
- MTP acceptance rate
- —
- TTFB
- 2.18 GB
- Power draw
- 60.7 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen2.5-Omni-3B
- Quantization
- IQ2_XS
- Framework
- llama.cpp
- Version
- v0.37
- Context length
- 2048 tokens
- Batch size
- 512
- GPU layers
- 28
- Flash Attention
- On
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)
- 5.12 GB
- OS
- Ubuntu 22.04
- Driver
- 565.57
- CUDA
- N/A
- Power draw
- 60.7 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.37 · Ubuntu 22.04 · CUDA N/A · 2048 ctx
273.09 tok/s
Decode
1,000.97 tok/s
Prefill
1.79 s
TTFT
5.12 GB
VRAM
—
MTP
60.7 W
Power draw
L2 batch seed 2026-10-16: llama.cpp on hw_id=88 running model_id=60 at IQ2_XS