Qwen2.5-Omni-3B · IQ2_XS
This page aggregates 2 real-world runs of Qwen2.5-Omni-3B (IQ2_XS) on Intel Arc A770 16GB with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
113.09 tok/s
Decode
Decode speed
493.14 tok/s
Prefill
Prefill speed
0.86 s
TTFT
Time to first token
5.12 GB
VRAM
VRAM usage
2
Measured runs
1
Independent sources
Other
Source platforms
2 days ago
Last verified
Performance
- llama.cpp · IQ2_XS (current)Decode 113.09 · Prefill 493.14 ·
- vLLM · IQ2_M Decode 105.24 · Prefill 529.61 ·
- vLLM · IQ4_XS Decode 86.78 · Prefill 278.58 ·
- llama.cpp · Q3_K_S Decode 92.22 · Prefill 373.8 ·
- Strata · IQ2_M Decode 145.2 · Prefill 469.28 ·
Core figures
- Decode (avg)
- 113.09 tok/s
- Prefill (avg)
- 493.14 tok/s
- TTFT (avg)
- 0.86 s
- VRAM (avg)
- 5.12 GB
- MTP acceptance rate
- —
- TTFB
- 0.9 GB
- Power draw
- 221 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen2.5-Omni-3B
- Quantization
- IQ2_XS
- Framework
- llama.cpp
- Version
- v0.51
- Context length
- 4096 tokens
- Batch size
- 512
- GPU layers
- 99
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- Intel Arc A770 16GB
- Nominal VRAM
- 16 GB
- Measured VRAM (avg)
- 5.12 GB
- OS
- Windows 11 23H2
- Driver
- Intel Arc 31.0.101.5186
- CUDA
- 12.6
- Power draw
- 217.5 W
Sources & evidence
2 measured records in total, each traceable to its original source
v0.51 · Windows 11 23H2 · CUDA 12.6 · 4096 ctx
119.33 tok/s
Decode
579.72 tok/s
Prefill
1.42 s
TTFT
5.12 GB
VRAM
—
MTP
217.5 W
Power draw
L2 batch seed 2026-10-16: llama.cpp on hw_id=97 running model_id=60 at IQ2_XS
v0.49 · Ubuntu 24.04 · CUDA 12.6 · 32768 ctx
106.85 tok/s
Decode
406.55 tok/s
Prefill
0.29 s
TTFT
5.12 GB
VRAM
—
MTP
224.5 W
Power draw
L2 batch seed 2026-10-13: llama.cpp on hw_id=97 running model_id=60 at IQ2_XS