Qwen-family LLMs are quietly becoming the backbone of modern audio models; One chart for the architectures of 100+ audio models [R]

I started mapping the building blocks shared across all the models in audio.cpp. The result ended up being more interesting than I expected.

Qwen has become by far the most common language backbone in this collection: 32 audio model families use a Qwen-family architecture, and 20 of them use Qwen3 LLM specifically.

And it’s no longer just TTS. Qwen-based models now show up across speech synthesis, ASR/audio understanding, music generation, speech-to-speech, and even audio/video models.

The 2nd chart, Task × Technology Matrix, shows which build blocks power which types of audio models.