Overview
Hy-SimulMT-30B-A3B-Interleave is Tencent Hunyuan’s open-weight model for low-latency simultaneous translation. It has approximately 30 billion parameters and is distributed under the Apache 2.0 license. (huggingface.co)
Capabilities and uses
- Translates streaming speech bidirectionally between Chinese and English as the speaker talks. The Interleave variant returns both source text and translated text. It is intended for settings such as meetings, speeches and interviews. (huggingface.co)
- Tencent’s SimulInfer service supports microphone input, audio file uploads and real-time WebSocket input. Speech output and voice cloning require a separately configured text-to-speech service; they are not standalone text outputs of this checkpoint. (github.com)
- In Tencent’s RealSI long-form evaluation, the model family achieved approximately 3 seconds of StreamLAAL. This is a streaming lag metric, not a guarantee of end-to-end latency or time to first speech output. (huggingface.co)
Limitations and deployment
The announced real-time translation directions are Chinese to English and English to Chinese. Translation quality and latency can vary with the audio and deployment. Tencent provides SimulInfer for inference and SimulTrain for continued training; self-hosting requires suitable computing resources and configuration. (huggingface.co)
Release date
October 26, 2026, according to Tencent’s model announcement. (huggingface.co)
Pricing
As of November 15, 2026, an official hosted API price specifically for this Interleave checkpoint could not be verified on Tencent Cloud’s domestic CNY or international USD pricing pages. Input and output charges are therefore unverified; no price or billing unit is asserted. Other Hy-MT or Hunyuan model prices should not be substituted. Self-hosting and any separate TTS service may incur infrastructure costs. (cloud.tencent.com)