Back to search results
Llama-3.2-3B-Instruct · Q2_K
Intel Arc B58012GBllama.cpp v0.38L2 Cross-framework verified
This page aggregates 1 real-world runs of Llama-3.2-3B-Instruct (Q2_K) on Intel Arc B580 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
169.62 tok/s
Decode
Decode speed
813.51 tok/s
Prefill
Prefill speed
2.46 s
TTFT
Time to first token
5.57 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
16 days ago
Last verified
Performance
- llama.cpp · Q2_K (current)Decode 169.62 · Prefill 813.51 ·
- llama.cpp · Q5_K_S Decode 120.67 · Prefill 490.78 ·
- Ollama · FP16 Decode 62.55 · Prefill 296.07 ·
- vLLM · IQ3_XS Decode 157.7 · Prefill 718.89 ·
- vLLM · Q4_K_M Decode 135.38 · Prefill 477.15 ·
- Strata · FP16 Decode 92 · Prefill 368.95 ·
- Strata · Q4_K_M Decode 163.66 · Prefill 710.01 ·
Core figures
- Decode (avg)
- 169.62 tok/s
- Prefill (avg)
- 813.51 tok/s
- TTFT (avg)
- 2.46 s
- VRAM (avg)
- 5.57 GB
- MTP acceptance rate
- —
- TTFB
- 1.36 GB
- Power draw
- 161 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Llama-3.2-3B-Instruct
- Quantization
- Q2_K
- Framework
- llama.cpp
- Version
- v0.38
- Context length
- 32768 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
- Intel Arc B580
- Nominal VRAM
- 12 GB
- Measured VRAM (avg)
- 5.57 GB
- OS
- Windows 11 24H2
- Driver
- Intel Arc 32.0.101.6552
- CUDA
- 12.4
- Power draw
- 161 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.38 · Windows 11 24H2 · CUDA 12.4 · 32768 ctx
169.62 tok/s
Decode
813.51 tok/s
Prefill
2.46 s
TTFT
5.57 GB
VRAM
—
MTP
161 W
Power draw
L2 batch seed 2026-10-19: llama.cpp on hw_id=94 running model_id=70 at Q2_K