What Qwen-Audio 3.1 cut, and by how much

Market · Voice AI · 23 September 2026

What Qwen-Audio 3.1 cut, and by how much

Reported reductions in Alibaba's published voice API list prices, by endpoint. Hover a bar for the detail.

ASRspeech recognition

up to 95%

Speech recognition: the steepest cut in the release, up to 95%. ASR-Next adds speaker separation, emotion and noise detection.

Realtimefull-duplex

~ 85%

Realtime conversation: down about 85%. Realtime Plus expands context to 262,144 tokens and adds function calling.

TTStext to speech

~ 70%

Text to speech: down about 70%. TTS-Next generates speech plus ambient audio in a single call.

All three cuts came in the same release, moving the price war that hit text models into audio.

The reductions are real, but none of the published material gives the old or new prices in currency — treat them as direction, not a budget line.

Source: Alibaba / Qwen-Audio 3.1 release, 23 September 2026, as reported by AI/TLDR, AI Weekly, byteiota and WION. Percentages are reported list-price reductions, not verified invoice savings.