NVIDIA · First MLX build 10 Feb 2026
diar_streaming_sortformer_4spk-v2.1
MLX
Audio
A 118M parameter voice activity detection model from NVIDIA, with 2 MLX builds, from fp16 to fp32.
- Parameters
- 118M From Hugging Face
- Quantisation
- fp16 diar_streaming_sortformer_4spk-v2.1-fp16
- Memory needed
- 3 GB Estimate, unified memory
- Download
- 236 MB Weights, estimated
- Downloads
- 1.0K This build, last 30 days
- Task
- Voice activity detection Audio
- MLX builds
- 2 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 3 GB at fp16
8 GB
16 GB
24 GB
32 GB
64 GB
128 GB
Fits with headroomTight, close other appsWon’t fit
Estimated from the size of the weights, plus a fifth for context and working memory and 2 GB for macOS and your apps. Long contexts need more.
Run it
Terminal · Hugging Face CLI$ pip install -U huggingface_hub
$ hf download mlx-community/diar_streaming_sortformer_4spk-v2.1-fp16MLX builds
2 on mlx-community| diar_streaming_sortformer_4spk-v2.1-fp16 | fp16 | 236 MB | 3 GB | 1.0K |
|---|---|---|---|---|
| diar_streaming_sortformer_4spk-v2.1-fp32 | fp32 | 472 MB | 3 GB | 228 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.