Release Notes

v0.1.1

First PyPI release. Install: uv pip install "sglang-omni==0.1.1".

Highlights since the early 0.1 line:

• TTS: dots.tts, ZONOS2; TTS architecture refactor (shared engine/ref-encode/vocoder paths); OpenAI speech + uploaded voices

• ASR: Fun-ASR, ARK-ASR; shared transcription serving; Qwen3-ASR pre-LM encoder and multilingual support

• Router: multiprocess CP/DP router; full TTS route set (/v1/audio/speech, batch, stream, voices)

• Runtime: CUDA-IPC weight sync / MPS-DP; speech API surface cleanup

• Omni: Qwen3-Omni talker/streaming fixes and perf work

See Installation. New model cookbooks are listed on the docs home page.