Qwen3.8-Omni-Flash
No published training cutoff
qwen3.8-omni-flash
$0.15 in, $0.47 out per 1M tokens
Qwen3.8-LiveTranslate (Flash Realtime)
Alibaba’s hosted real-time interpretation model, a competitor to gpt-realtime-translate and Gemini 3.5 Live Translate.
Model idqwen3.8-livetranslate-flash-realtime
Price$7.50 audio in, $20 text out per 1M tokens
Checked29 September 2026
Where it stands
22newer speech and audio models since its release
- NVIDIA Nemotron 3.5 Transcribe (30B-A4B)28 Oct 2026
- Hy-SimulMT-30B-A3B22 Oct 2026
- Eleven Realtime v1 Max / Mini20 Oct 2026
- Voxtral Mini Transcribe 3 (+ Mini Transcribe Realtime 3)20 Oct 2026
- Voxtral Small Transcribe 320 Oct 2026
The 5 newest are listed. All speech and audio models
- Before it from Alibaba
- Qwen3-ASR (0.6B / 1.7B) + Qwen3-ForcedAligner29 Jan 2026
- After it from Alibaba
- Qwen-Audio-3.1-TTS-Next22 Sep 2026
- Weights
- Closed weights; access is hosted only.
How to call it
2 ways to use Qwen3.8-LiveTranslate (Flash Realtime), with the model id each provider expects.
| Provider | Model id | Endpoint | Docs |
|---|---|---|---|
| QwenCloud (Realtime WebSocket) | qwen3.8-livetranslate-flash-realtime |
wss://maas.qwencloudapi.com/api-ws/v1/realtime?model=qwen3.8-livetranslate-flash-realtime |
qwencloud.com |
| Alibaba Cloud Model Studio | qwen3.8-livetranslate-flash-realtime |
alibabacloud.com |
Pricing
| Audio input | $7.50 per 1M tokens |
|---|---|
| Image input | $0.55 per 1M tokens |
| Text output | $20 per 1M tokens |
| Audio output | $30 per 1M tokens |
As published: “USD per 1M tokens (QwenCloud list price; press estimates about $1.54 per hour of speech in and out)”.
Source: qwencloud.com, checked 29 September 2026.
Specs
- Input
- Audio, images and text
- Output
- Audio and text
- Context window
- 53,248 tokens
- Max output
- 4,096 tokens
- Open weights
- No
- Licence
- Proprietary
- Released
- 19 September 2026
- Status
- Current
- Type
- Speech and audio
What stands out
What its maker marketed at launch, and what people found later, each with its source.
Simultaneous interpretation with lower lag
Streams translated speech and text while the speaker is still talking; average lagging (LAAL) cut from 2.8 s to 2.3 s across 60 languages with a new ‘Interleave’ architecture.
Multi-speaker diarization with per-speaker voice cloning
Tells speakers apart in multi-party speech and keeps each speaker’s own voice in the translated audio; synchronized bilingual on-screen display.
Long-context disambiguation
Uses conversation history to keep names and terminology consistent across a session.
Notes
Understands 60 languages and speaks 29 (the rest get text-only translation). Thinker-talker hybrid MoE on the Qwen-Omni stack (press). API-only, no open weights and no announced timeline for them. MindStudio’s hands-on found short sentences fine but weak end-of-turn detection, so developers need their own turn-taking logic. Announced on X 2026-09-19 (294k views by 2026-09-29), shortly before Apsara 2026.
Sources: https://x.com/Alibaba_Qwen/status/2101206705111757253 · https://qwen.ai/blog?id=qwen3.8-livetranslate · https://www.qwencloud.com/models/qwen3.8-livetranslate-flash-realtime · https://www.marktechpost.com/2026/09/19/alibaba-qwen-team-releases-qwen3-8-livetranslate/ · https://www.mindstudio.ai/blog/qwen3-8-livetranslate-hands-on