Back to search results
Qwen2.5-Omni-3B · Q4_K_M
Apple M2 UltravLLM v0.35L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen2.5-Omni-3B (Q4_K_M) on Apple M2 Ultra with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
157.8 tok/s
Decode
Decode speed
762.2 tok/s
Prefill
Prefill speed
2.49 s
TTFT
Time to first token
7.27 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 Decode 215.16 · Prefill 690.54 ·
- llama.cpp · Q3_K_M Decode 171.07 · Prefill 900.09 ·
- vLLM · Q4_K_M (current)Decode 157.8 · Prefill 762.2 ·
Core figures
- Decode (avg)
- 157.8 tok/s
- Prefill (avg)
- 762.2 tok/s
- TTFT (avg)
- 2.49 s
- VRAM (avg)
- 7.27 GB
- MTP acceptance rate
- —
- TTFB
- 1.26 GB
- Power draw
- 65.3 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen2.5-Omni-3B
- Quantization
- Q4_K_M
- Framework
- vLLM
- Version
- v0.35
- Context length
- 2048 tokens
- Batch size
- 16
- GPU layers
- 999
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- Apple M2 Ultra
- Measured VRAM (avg)
- 7.27 GB
- OS
- macOS 15.0
- Driver
- Apple M2 Ultra
- CUDA
- N/A
- Power draw
- 65.3 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.35 · macOS 15.0 · CUDA N/A · 2048 ctx
157.8 tok/s
Decode
762.2 tok/s
Prefill
2.49 s
TTFT
7.27 GB
VRAM
—
MTP
65.3 W
Power draw
L2 batch seed 2026-10-16: vLLM on hw_id=103 running model_id=60 at Q4_K_M