Fun ASR Nano needs about 1.0 GB of memory at its main file (a 0.5 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 4 quants, from 0.5 GB (F16) to 0.8 GB (Q8_0).
Fun ASR Nano – small, efficient ASR model for CPU/edge
Fun ASR Nano is a 0.234 billion‑parameter encoder with a Qwen3‑0.6B LLM decoder. It runs on the FunASR llama.cpp runtime and delivers 8.30 % CER on a Mandarin benchmark. The model is released under Apache‑2.0.
Good at
- Low‑latency speech recognition on CPUs
- Small model size for edge devices
- High accuracy with Qwen3 decoder
Limits
- Requires both encoder and LLM GGUF files
- No vision support
- Limited to 8 k context
Best for
- Embedded speech‑to‑text applications
- Edge devices with limited VRAM
- Rapid prototyping of ASR pipelines
Who it’s for. Developers building lightweight ASR solutions on CPUs or edge hardware. The model is suitable for projects needing accurate Mandarin transcription without GPU resources.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| F16 | 0.5 GB | 1.0 GB | Fits | Get file → | |
| — | 0.5 GB | 1.0 GB | Fits | Get file → | |
| — | 0.6 GB | 1.1 GB | Fits | Get file → | |
| Q8_0 | 0.8 GB | 1.4 GB | Fits | Get file → |
For 8 GB VRAM, use the Q8_0 quantization (0.8 GB file, ~1.4 GB memory). If memory is tighter, choose the unknown quant (0.48 GB file, ~1 GB memory). The F16 quant (0.47 GB file) also fits within 1 GB memory and offers good speed.
What GPU do you need for Fun ASR Nano?
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 Q8_0 (0.8 GB file, about 1.4 GB in use).
Which Fun ASR Nano quant should you choose?
If memory is tight, the smallest file is F16 at 0.5 GB, with the quality trade-offs that come with fewer bits. The unquantized model. It gives original quality and needs the most memory, so it is mainly for conversion or reference.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Fun ASR Nano
How much VRAM does Fun ASR Nano need?
About 1.0 GB at its main file with an 8K context window. That is the 0.5 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 Fun ASR Nano on an 8 GB GPU?
Yes, with Q8_0: the 0.8 GB file needs about 1.4 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Fun ASR Nano GGUF?
It depends on the quant: F16 is 0.5 GB, file is 0.5 GB, file is 0.6 GB, Q8_0 is 0.8 GB.
Where can I download Fun ASR Nano GGUF?
The files are on Hugging Face in the FunAudioLLM/Fun-ASR-Nano-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 Fun ASR Nano use?
Hugging Face lists the license as apache-2.0. 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 Fun ASR Nano on 8 GB VRAM?
Yes, the Q8_0 quant uses 0.8 GB file size and requires about 1.4 GB memory at 8 k context, fitting within 8 GB VRAM.
Does Fun ASR Nano support languages other than Mandarin?
The model’s benchmark is Mandarin; language support beyond that is not specified.
Is Fun ASR Nano compatible with GPU acceleration?
It is designed for the FunASR llama.cpp runtime, which is CPU/edge focused and does not require GPU.