Back to search results
Qwen2.5-0.5B-Instruct · Q6_K
AMD Strix Halo (Radeon 8060S)llama.cpp v0.23L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen2.5-0.5B-Instruct (Q6_K) on AMD Strix Halo (Radeon 8060S) with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
171.39 tok/s
Decode
Decode speed
894.49 tok/s
Prefill
Prefill speed
1.55 s
TTFT
Time to first token
3.15 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 · FP16 Decode 124.48 · Prefill 656.36 ·
- llama.cpp · Q6_K (current)Decode 171.39 · Prefill 894.49 ·
- vLLM · Q5_K_S Decode 228.54 · Prefill 950.08 ·
Core figures
- Decode (avg)
- 171.39 tok/s
- Prefill (avg)
- 894.49 tok/s
- TTFT (avg)
- 1.55 s
- VRAM (avg)
- 3.15 GB
- MTP acceptance rate
- —
- TTFB
- 1.37 GB
- Power draw
- 45.2 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen2.5-0.5B-Instruct
- Quantization
- Q6_K
- Framework
- llama.cpp
- Version
- v0.23
- Context length
- 4096 tokens
- Batch size
- 256
- 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
- AMD Strix Halo (Radeon 8060S)
- Measured VRAM (avg)
- 3.15 GB
- OS
- Ubuntu 22.04
- Driver
- amdgpu 6.11
- CUDA
- N/A
- Power draw
- 45.2 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.23 · Ubuntu 22.04 · CUDA N/A · 4096 ctx
171.39 tok/s
Decode
894.49 tok/s
Prefill
1.55 s
TTFT
3.15 GB
VRAM
—
MTP
45.2 W
Power draw
L2 batch seed 2026-10-16: llama.cpp on hw_id=87 running model_id=59 at Q6_K