SenseVoiceSmall needs about 0.8 GB of memory at Q8_0 (a 0.3 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 3 quants, from 0.1 GB (Q4_K) to 0.5 GB (F16).
SenseVoiceSmall – a lightweight, non‑autoregressive ASR model with multilingual, emotion, and event detection
SenseVoiceSmall is a 0.234 billion‑parameter encoder‑only model designed for efficient speech understanding. It supports automatic speech recognition, spoken language identification, speech emotion recognition, and audio event detection in over 50 languages. The model runs as a single binary via llama.cpp / GGUF, with built‑in VAD and low inference latency.
Good at
- Fast non‑autoregressive inference
- Multilingual ASR with native language ID
- Rich emotion and audio‑event detection
Limits
- No GPU required, but large VRAM needed for higher quants
- No vision or text‑generation capabilities
- License is other – check usage terms
Best for
- Edge devices needing fast speech transcription
- Multilingual voice‑assistant back‑ends
- Audio‑event monitoring in real‑time
Who it’s for. Developers building speech‑enabled applications that require quick, accurate transcription and contextual tags without heavy GPU resources. Ideal for embedded systems, voice‑assistant services, and audio‑analysis pipelines.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K | 0.1 GB | 0.7 GB | Fits | Get file → | |
| Q8_0 | 0.3 GB | 0.8 GB | Fits | Get file → | |
| F16 | 0.5 GB | 1.0 GB | Fits | Get file → |
For 8 GB VRAM, the Q4_K quant (0.14 GB file, ~0.7 GB VRAM at 8k context) is recommended. If more VRAM is available, Q8_0 (0.25 GB, ~0.8 GB VRAM) offers higher precision. The F16 quant (0.47 GB, ~1 GB VRAM) requires the most memory and is best for systems with ample VRAM.
What GPU do you need for SenseVoiceSmall?
These figures add the file size, an fp16 KV cache at an 8K context window and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is F16 (0.5 GB file, about 1.0 GB in use).
Which SenseVoiceSmall quant should you choose?
Q8_0 (0.3 GB) is the middle option in this repository. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.
If memory is tight, the smallest file is Q4_K at 0.1 GB, with the quality trade-offs that come with fewer bits. The usual sweet spot between file size and quality, and the quant most people start with.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about SenseVoiceSmall
How much VRAM does SenseVoiceSmall need?
About 0.8 GB at Q8_0 with an 8K context window. That is the 0.3 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run SenseVoiceSmall on an 8 GB GPU?
Yes, with F16: the 0.5 GB file needs about 1.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of SenseVoiceSmall GGUF?
It depends on the quant: Q4_K is 0.1 GB, Q8_0 is 0.3 GB, F16 is 0.5 GB.
Where can I download SenseVoiceSmall GGUF?
The files are on Hugging Face in the cstr/sensevoice-small-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does SenseVoiceSmall use?
Hugging Face lists the license as other. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run SenseVoiceSmall on 8 GB VRAM?
Yes, the Q4_K quant uses about 0.7 GB VRAM at 8k context, fitting comfortably in 8 GB.
Does SenseVoiceSmall support vision tasks?
No, it is an audio‑only model with no vision or image‑processing capabilities.
What languages does SenseVoiceSmall cover?
It supports over 50 languages, providing native spoken language identification.