audio model · LFM · Linux
Can I run LFM3 Audio 2B on Nvidia DGX Spark (128GB)?
Yes. LFM3 Audio 2B runs on Nvidia DGX Spark (128GB) at Q8_0, audio.cpp ASR (~4.61 GB of ~112 GB usable).
Runs at Q8_0, audio.cpp ASR using ~4.61 GB of ~112 GB usable.
- Peak memory
- ~4.61 GB
- Usable on device
- ~112 GB
- Device memory
- 128 GB
- Quant
- Q8_0, audio.cpp ASR
How to run it
Use audio.cpp (experimental) at Q8_0, audio.cpp ASR. It loads as a single model and runs on a GPU.
- Type
- Speech recognition
- Parameters
- 2B
- Peak memory
- ~4.61 GB at Q8_0, audio.cpp ASR
- License
- LFM Open License v1.0
- Memory
- 128 GB unified
- Usable for weights
- ~112 GB
- Power draw
- ~140 W
- Best runtime
- llama.cpp (CUDA) / Ollama (CUDA)
You could also run
Run LFM3 Audio 2B on other hardware
FAQ
Can Nvidia DGX Spark (128GB) run LFM3 Audio 2B?
Yes. LFM3 Audio 2B runs on Nvidia DGX Spark (128GB) at Q8_0, audio.cpp ASR (~4.61 GB of ~112 GB usable).
How much memory does LFM3 Audio 2B need?
Nvidia DGX Spark (128GB) has room to spare. At Q8_0, audio.cpp ASR the realistic peak is ~4.61 GB of memory.
What do I use to run LFM3 Audio 2B locally?
LFM3 Audio 2B runs in audio.cpp (experimental). It needs a GPU.
Sources
-
docs.nvidia.com · 2 sources
-
github.com · 2 sources
-
nvidia.com · 1 source
-
nvidianews.nvidia.com · 1 source
-
ollama.com · 2 sources
-
tomshardware.com · 1 source
VRAM figures are sourced peak-usage anchors at the noted quant; catalog updated 2026-11-01. See methodology.