FU

Fun-ASR-Nano-2512-hf-854d88f

FunGGUF by FunAudioLLMUpdated 2026-07-29
Download on Hugging Face
Parameters—
Context—
Layers—
Architectureaudiocpp
Smallest quant1.0 GB
Quick answer

Fun-ASR-Nano-2512-hf-854d88f needs about 1.7 GB of memory at Q8_0 (a 1.0 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 2 quants, from 1.0 GB (Q8_0) to 1.7 GB (F16).

Fun-ASR-Nano-2512: lightweight multilingual speech recognition

Fun-ASR-Nano-2512 is an end‑to‑end ASR model supporting Chinese, English and Japanese. It is released in GGUF format for audio.cpp and can be run on CPU or CUDA. The model is trained on tens of millions of hours of real speech data.

Good at

  • Multilingual transcription of Chinese, English, Japanese
  • Low‑latency inference on modest hardware
  • Supports dialect and accent variations

Limits

  • No streaming or timestamp output
  • Requires 8k context for memory estimates
  • License is FunASR Model Open Source License v1.1

Best for

  • Transcribing short audio clips
  • Batch processing of recorded interviews
  • Local transcription on edge devices

Who it’s for. Developers building offline speech‑to‑text pipelines that need a small, multilingual model. It is suitable for applications where GPU memory is limited and streaming is not required.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q8_01.0 GB 1.7 GB
Fits Get file →
F161.7 GB 2.4 GB
Fits Get file →

For 8 GB VRAM, use the Q8_0 quantization (1.05 GB file, ~1.7 GB VRAM at 8k context). If more VRAM is available, the F16 quant (1.68 GB file, ~2.4 GB VRAM) offers higher precision. Memory estimates are approximate.

What GPU do you need for Fun-ASR-Nano-2512-hf-854d88f?

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 (1.7 GB file, about 2.4 GB in use).

Which Fun-ASR-Nano-2512-hf-854d88f quant should you choose?

Q8_0 (1.0 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.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureaudiocpp

Source

Quantized byFunAudioLLM
FamilyFun
Licenseother
Downloads (HF)69,312
Last updated2026-07-29

Frequently asked questions about Fun-ASR-Nano-2512-hf-854d88f

How much VRAM does Fun-ASR-Nano-2512-hf-854d88f need?

About 1.7 GB at Q8_0 with an 8K context window. That is the 1.0 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-2512-hf-854d88f on an 8 GB GPU?

Yes, with F16: the 1.7 GB file needs about 2.4 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Fun-ASR-Nano-2512-hf-854d88f GGUF?

It depends on the quant: Q8_0 is 1.0 GB, F16 is 1.7 GB.

Where can I download Fun-ASR-Nano-2512-hf-854d88f GGUF?

The files are on Hugging Face in the FunAudioLLM/Fun-ASR-Nano-2512-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-2512-hf-854d88f 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 Fun-ASR-Nano on 8 GB VRAM?

Yes, the Q8_0 quantization requires about 1.7 GB VRAM at 8k context, which fits in 8 GB.

Does the model support streaming inference?

No, Fun-ASR-Nano does not expose streaming or timestamp output.

What languages can it transcribe?

It supports Chinese, English, and Japanese, with coverage of 7 Chinese dialects and 26 regional accents.

Other GGUFs of Fun-ASR-Nano-2512-hf

More from FunAudioLLM

All by FunAudioLLM →

More Fun models

All Fun →