Updated

Qwen3.8-LiveTranslate

Alibaba debuted Qwen3.8-LiveTranslate at Apsara 2026, a simultaneous interpretation model for real-time translation designed for smoother, more responsive live translation. It launched alongside the Qwen-Audio 3.1 speech and image model updates.

By No Code MBA · Facts checked

Provider
Qwen
Announcement date
Status at last check
Generally available
Model family
Qwen3.8
Preceding model
Qwen3.5-LiveTranslate
Capabilities
Audio, Text, Image

What changed

Qwen3.8-LiveTranslate is a simultaneous interpretation model that takes audio and images and outputs translated text and speech. Model Studio docs list understanding of 60 languages and speech output in 29, with automatic source language detection, unlike the 3.5 version. Alibaba says it cuts average lag (LAAL) by nearly 20%, from 2.8 to 2.3 seconds, and improves fidelity, fluency and brevity. It can clone the speaker's voice during translation without an extra fee.

Access and availability

Connect to the Alibaba Cloud Model Studio WebSocket Realtime API with model qwen3.8-livetranslate-flash-realtime, in the Singapore or China (Beijing) region. Weights are not released.

Official access documentation

Pricing

Singapore region: $7.50 per million audio input tokens, $0.55 per million image input tokens, $20 per million text output tokens and $30 per million audio output tokens

Official pricing

What you can build

Test it against a live bilingual call or meeting feed and compare latency and fluency with a streaming translation API before committing.

Limitations

Context is capped at 53,248 tokens (4,096 output), and Singapore rate limits are 10 requests and 100,000 tokens per minute. Function calling, structured outputs, web search, context caching, batch inference and fine-tuning are not supported.

Sources and release history