Alibaba (Qwen)Notable

Qwen-Audio-3.0-Realtime

Release date unpublished

Alibaba's previous-generation full-duplex realtime voice model family (Plus and Flash) on Model Studio. It is still listed, but no longer recommended.

Price per minute: —
Pricing: qwen-audio-3.0-realtime-plus, Singapore, per 1M tokens (shown in CNY): text in 5.995, audio in 47.963, text out 47.963, audio out 179.861 (text output not charged when audio returned). Flash pricing not fetched.
Inputs: audio, text
Tool calling: Yes
Languages: 11
Voices: 5
Big Bench Audio: 99.2%
Access: Generally available
Licence: Proprietary

Strengths & weaknesses

Strengths

Big Bench Audio 99.2% on Plus, #2 on AA
Function calling and web search
11 languages plus 21 Chinese varieties; 5 system voices plus cloning

Weaknesses

Superseded by Qwen-Audio-3.1-Realtime and no longer recommended in Alibaba's docs
40,960-token context with 8,192 max output on Plus
AA S2S Index 69.5, far below its reasoning score

Notable: #2 on AA Big Bench Audio (Plus 99.2%); Flash scores 96.1% (#13); S2S Index Plus 69.5, Flash 67.7. releaseDate is null because no Alibaba page gives one; Chinese media (ithome, 10jqka, qbitai) report 2026-07-15. Alibaba's speech-to-speech page labels it 'earlier generation' and not recommended, so drop it if superseded models are out of scope. Pricing below was published in CNY on the Singapore page.

Links

API model strings

alibabaqwen-audio-3.0-realtime-plus
alibabaqwen-audio-3.0-realtime-flash