Back to search results
Llama-3.2-3B-Instruct · IQ1_S
NVIDIA RTX 508016GBllama.cpp v0.38L2 Cross-framework verified
This page aggregates 1 real-world runs of Llama-3.2-3B-Instruct (IQ1_S) on NVIDIA RTX 5080 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
351.28 tok/s
Decode
Decode speed
1,853.09 tok/s
Prefill
Prefill speed
1.2 s
TTFT
Time to first token
4.75 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
Today
Last verified
Performance
- llama.cpp · IQ1_S (current)Decode 351.28 · Prefill 1853.09 ·
- Ollama · IQ4_XS Decode 362.6 · Prefill 1325.05 ·
Core figures
- Decode (avg)
- 351.28 tok/s
- Prefill (avg)
- 1,853.09 tok/s
- TTFT (avg)
- 1.2 s
- VRAM (avg)
- 4.75 GB
- MTP acceptance rate
- —
- TTFB
- 2.81 GB
- Power draw
- 227.4 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Llama-3.2-3B-Instruct
- Quantization
- IQ1_S
- Framework
- llama.cpp
- Version
- v0.38
- Context length
- 2048 tokens
- Batch size
- 256
- GPU layers
- 20
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- NVIDIA RTX 5080
- Nominal VRAM
- 16 GB
- Measured VRAM (avg)
- 4.75 GB
- OS
- Windows 11 24H2
- Driver
- 572.83
- CUDA
- 13.0
- Power draw
- 227.4 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.38 · Windows 11 24H2 · CUDA 13.0 · 2048 ctx
351.28 tok/s
Decode
1,853.09 tok/s
Prefill
1.2 s
TTFT
4.75 GB
VRAM
—
MTP
227.4 W
Power draw
L2 batch seed 2026-10-18: llama.cpp on hw_id=14 running model_id=70 at IQ1_S