FunASR llama.cpp runtime v0.1.9

Prebuilt self-contained binaries for the FunASR llama.cpp / GGUF runtime, with Fun-ASR-Nano, SenseVoice, and Paraformer support plus built-in FSMN-VAD.

This release updates the Fun-ASR Latest Release from runtime v0.1.4 to v0.1.9 and adds the previously missing Vulkan and Windows CUDA packages. Its source snapshot also contains the corrected bilingual vLLM guide: users now clone the canonical modelscope/FunASR service scripts before running serve_vllm.py or serve_realtime_ws.py.

Choose an asset

• Default linux-x64 / windows-x64: maximum CPU compatibility.
• linux-x64-avx2 / windows-x64-avx2: higher throughput on CPUs with AVX2, FMA, F16C, and BMI2.
• linux-x64-vulkan / windows-x64-vulkan: Vulkan backend; requires a working Vulkan driver and ICD.
• windows-x64-cuda: CUDA backend targeting architecture 86; build from source for other GPU architectures.
• Native linux-arm64 and macos-arm64 packages are included.

Download a default quantized model with:

pip install -U huggingface_hub
bash download-funasr-model.sh nano

Then run llama-funasr-cli; no Python ASR runtime or local compilation is required.

All nine files are byte-for-byte identical to the digest-verified assets from modelscope/FunASR runtime v0.1.9.

Documentation: https://github.com/QwenAudio/Fun-ASR/blob/runtime-llamacpp-v0.1.9/runtime/llama.cpp/README.md

FunASR llama.cpp runtime v0.1.4

Prebuilt self-contained binaries for the FunASR llama.cpp / GGUF runtime.

This release keeps the default x64 assets conservative for maximum CPU compatibility and adds explicit x64-avx2 assets for CPUs with AVX2/FMA/F16C/BMI2. If v0.1.3 fixed SIGILL but was too slow on your machine, try funasr-llamacpp-windows-x64-avx2.zip.

Use the default x64 package if the AVX2 package fails with Illegal instruction / SIGILL.

FunASR llama.cpp runtime v0.1.3

Portable rebuild of the FunASR llama.cpp / GGUF runtime binaries.

This release rebuilds the v0.1.2 runtime with conservative CPU flags for the generic x64 assets. In particular, it disables ggml native CPU tuning and high-ISA AVX/AVX2/AVX512/VNNI/FMA/F16C/BMI2 flags so the prebuilt binaries do not depend on the GitHub Actions runner CPU instruction set.

Use this version if v0.1.2 crashes with Illegal instruction, SIGILL, or Windows error 0xC000001D.

FunASR llama.cpp runtime v0.1.2

Prebuilt self-contained binaries for running Fun-ASR-Nano (and SenseVoice / Paraformer) locally with the FunASR llama.cpp / GGUF runtime — built-in FSMN-VAD, whisper.cpp-style on-device ASR, strong on Chinese.

New: q8 GGUF models are ~half the size of f16 with the same accuracy.

bash download-funasr-model.sh nano ./gguf
llama-funasr-cli --enc ./gguf/funasr-encoder-f16.gguf -m ./gguf/qwen3-0.6b-q8_0.gguf -a audio.wav --vad ./gguf/fsmn-vad.gguf

No Python, no build. Linux (x64/arm64), macOS (arm64), Windows (x64). Docs: runtime/llama.cpp/README.md.

FunASR llama.cpp runtime v0.1.1

Prebuilt, self-contained binaries to run Fun-ASR-Nano (and SenseVoice / Paraformer) locally with the FunASR llama.cpp / GGUF runtime — built-in FSMN-VAD, whisper.cpp-style on-device ASR, strong on Chinese.

bash download-funasr-model.sh nano ./gguf
llama-funasr-cli --enc ./gguf/funasr-encoder-f16.gguf -m ./gguf/qwen3-0.6b-q8_0.gguf -a audio.wav --vad ./gguf/fsmn-vad.gguf

No Python, no build, no torch. Binaries for Linux (x64/arm64), macOS (arm64), Windows (x64). Docs: runtime/llama.cpp/README.md. CER (micro-avg): Fun-ASR-Nano 8.30% vs whisper.cpp small 22%.

v1.0.0: Fun-ASR-Nano and Fun-ASR-MLT-Nano

Fun-ASR v1.0.0

The first official release of the Fun-ASR model family, trained on large-scale real speech data for low-latency speech recognition.

Checkpoints

Model	Supported languages	Parameters	Download
Fun-ASR-Nano-2512	Chinese, English, and Japanese. Chinese coverage includes 7 dialect groups and 26 regional accents.	800M	ModelScope / Hugging Face
Fun-ASR-MLT-Nano-2512	31 languages	800M	ModelScope / Hugging Face

Highlights

• Low-latency transcription with native punctuation.
• Chinese dialect and regional-accent recognition in the flagship Nano checkpoint.
• A separate MLT checkpoint for 31-language recognition.
• Hotword support for domain-specific terms.
• Speech and lyric recognition under music-background interference.
• FunASR and vLLM inference paths.

Capability boundary

Fun-ASR-Nano-2512 does not currently return reliable native character or word timestamps and does not provide speaker diarization by itself. Workflows that need VAD segmentation, punctuation, or speaker diarization should compose the relevant FunASR pipeline components. Use a timestamp-capable model such as Paraformer when precise text-aligned clipping is required.

Quick start

from funasr import AutoModel

model = AutoModel(
    model="FunAudioLLM/Fun-ASR-Nano-2512",
    trust_remote_code=True,
    device="cuda:0",
    hub="hf",
)
result = model.generate(input=["audio.wav"], cache={}, batch_size=1)
print(result[0]["text"])

Links

• Documentation
• vLLM guide
• FunASR toolkit