Qwen3-ASR-0.6B-chatllm-quantized

Available Models

• qwen3-asr-0.6b-q4_0.bin: 4-bit quantization (Decent accuracy, fastest inference). Recommended for free-tier CPU instances.
• qwen3-asr-0.6b-q8_0.bin: 8-bit quantization (High accuracy, slightly slower than Q4).

Usage with chatllm.cpp

1. Clone or download the binaries.
2. Run with the following command (requires chatllm-main):

Credits

• Original Model: Qwen Team (Alibaba Cloud)
• C++ Backend: foldl/chatllm.cpp
• Quantization: Pre-quantized using chatllm.cpp conversion scripts.

License