runtime-llamacpp-v0.1.6
modelscope/FunASRruntime-llamacpp-v0.1.6Jul 15, 2026by github-actions[bot]
AI Summary
This release introduces prebuilt self-contained binaries for the FunASR llama.cpp runtime, supporting SenseVoice, Paraformer, and Fun-ASR-Nano models with built-in FSMN-VAD. It eliminates the need for a Python runtime or local build by providing easy-to-use command-line tools. Users can download quantized models via a helper script and choose between standard x64 or optimized AVX2 assets for CPU compatibility and throughput.
Key Highlights
- Prebuilt self-contained binaries for llama.cpp/GGUF runtime.
- Support for SenseVoice, Paraformer, and Fun-ASR-Nano models with built-in FSMN-VAD.
- No Python runtime or local build required.
- Easy model download via helper script (`download-funasr-model.sh`).
- Optimized CPU assets (x64 vs x64-avx2) for higher throughput.
New Features
- Self-contained binary distributions
- Support for SenseVoice, Paraformer, and Fun-ASR-Nano models
- Built-in FSMN-VAD functionality
- Command-line interface tools (`llama-funasr-cli`, etc.)
- Model download helper script
Full Release Notes
Prebuilt self-contained binaries for the FunASR llama.cpp / GGUF runtime: SenseVoice, Paraformer and Fun-ASR-Nano with built-in FSMN-VAD. Download the default quantized model with `bash download-funasr-model.sh <sensevoice|paraformer|nano>` (the helper requires the Hugging Face CLI: `pip install -U huggingface_hub`), then run `llama-funasr-cli` / `llama-funasr-sensevoice` / `llama-funasr-paraformer`. Use the default x64 asset for maximum CPU compatibility; use the x64-avx2 asset on CPUs with AVX2/FMA/F16C/BMI2 for higher throughput. No Python ASR runtime or local build is required. Docs: https://github.com/modelscope/FunASR/blob/runtime-llamacpp-v0.1.6/runtime/llama.cpp/README.md