Tencent open-sources Hy-SimulMT, a 30B-A3B Chinese–English simultaneous speech interpreter with about 3 s lag
Confirmed
The takeaway
Tencent reports a StreamLAAL latency of about 3 seconds on its long-form RealSI evaluation, the lowest in its comparison, with translation quality above the open models and four of five commercial APIs it tested.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 2 of 5
- Last verified
- 27 October 2026
Your AI and this story
- GPT-6 Astra180 days after its cutoff
- Claude Opus 5.5119 days after its cutoff
- Gemini 4 Hydrogen210 days after its cutoff
- Grok 4.7149 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 119 days before it.
Key facts
- Weights: tencent/Hy-SimulMT-30B-A3B plus an ‘Interleave’ variant, FP8 and GGUF builds (Hugging Face repos created Oct 22, 2026; licence apache-2.0)
- Function (model card): bidirectional Chinese–English speech-to-speech simultaneous translation; continuous audio input, translated text and speech output; speech output and voice cloning need a configured TTS service
- Tencent’s numbers (model card): StreamLAAL ~3 s on RealSI long-form audio, lowest among systems shown; GEMBA quality above compared open models and four of five commercial APIs (vendor evaluation, not independent)
- Tooling: GitHub Tencent-Hunyuan/Hy-SimulMT with the SimulInfer inference service and SimulTrain training tools; also in Tencent’s Hy Translate app
What happened
Open simultaneous interpretation has lagged closed systems (ByteDance’s Seed LiveInterpret, OpenAI’s and Google’s live translation). Hy-SimulMT is a small-active-parameter open model aimed at meetings and speeches, with inference and training code released.
Sources
4 sources from 4 sites. Numbers match the chips in the text.
4 sources: 2 primary, 2 press
Primary
- Hugging Face: tencent/Hy-SimulMT-30B-A3B (model card)huggingface.co, official
- GitHub: Tencent-Hunyuan/Hy-SimulMTgithub.com, code
Press
- Pandaily: Tencent Hunyuan open-sources Hy-SimulMTpandaily.com, press
- TechNode: Tencent open-sources AI model for real-time Chinese-English interpretationtechnode.com, press
Changes
- Filed (model card read directly)