Primary use stated or inferred cautiously from official documentation.
T
TENCENT / tencent/Hy-SimulMT-30B-A3B-Interleave
Hy-SimulMT-30B-A3B-Interleave
Tencent Hunyuan checkpoint for simultaneous Chinese-English speech translation, included in the October 26, 2026 announcement. It receives incremental audio and produces translated text. The Interleave variant also retains source text; this is a different output, not a quantization. Repository creation on October 22 is not presented as the release date.
01 / SUMMARY
The essential
Maximum input capacity when published by the source.
Documented maximum generation limit.
Verified availability channels.
Primary functional family and verified specialties.
Declared terms for API access, model weights, or self-hosting.
02 / TECHNICAL SHEET
Limits and integration
| API ID | tencent/Hy-SimulMT-30B-A3B-Interleave |
|---|---|
| Model type | Voice and audio |
| Access model | Open weights |
| License | Apache 2.0 |
| Deployment | Self-hosting with vLLM and SimulInfer |
| Release | 26 OCT 2026 |
| Knowledge cutoff | Not published |
| Entrance | Audio |
| Exit | Text |
| Context window | Not published |
| maximum output | Not published |
| Reasoning | Not published |
| Published tools | |
| Structured Outings | Not published |
| Batch processing | Not published |
| Prompt cache | Not published |
| Fine-tuning | Not published |
| Verified platforms | Hugging Face · SimulInfer · vLLM |
Meetings, interviews and speeches where translation should continue as audio arrives. SimulInfer provides file uploads, WebSocket and a web interface; SimulTrain supports continued training. Speech output and voice cloning require a separately configured TTS service: they are not native audio outputs guaranteed by this checkpoint.
The announcement verifies bidirectional Chinese-English translation; the metadata language list does not demonstrate speech translation in all those languages. Recorded demonstrations do not measure live latency. StreamLAAL measures translation lag, not time to first audio. For llama.cpp support, the project recommends waiting for a compatible release; we do not present it as available stable support.
03 / CAPABILITIES
What it can do
Orientation
Meetings, interviews and speeches where translation should continue as audio arrives. SimulInfer provides file uploads, WebSocket and a web interface; SimulTrain supports continued training. Speech output and voice cloning require a separately configured TTS service: they are not native audio outputs guaranteed by this checkpoint.
Context
Not published · Not published
Tools and integration
Not published
Access
Hugging Face
Input modalities
Audio
04 / PRICES
Documented cost
| Concept | Worth | Unit/condition |
|---|---|---|
| Standard input | No published official API rate | Standard API rate |
| Cached input | Not published | Reading reused prefixes |
| Cache write or storage | Not published | The condition varies by provider |
| Standard output | Self-hosted infrastructure cost | May include reasoning tokens |
| Batch input | Not published | Asynchronous processing |
| Batch output | Not published | Asynchronous processing |
Weights under Apache 2.0, with FP8 and GGUF variants linked as formats of the same family. No hosted API price or universal context or duration limit has been verified. Operation requires resources for the model, audio transport and optional TTS. Voice cloning requires consent and rights to the voice used.
i Prices change and may depend on level, region or context length. Check the source before making a decision.
05 / EVALUATIONS
How to read the results
“
Tencent reports family-level Hy-SimulMT scores of 90.3 on FLEURS, 75.5 on CoVoST2, 81.2 on ACL6060 and 84.7 on RealSI-Sentence, plus approximately three seconds of StreamLAAL on long-form RealSI audio. These tables do not identify an independent score for each checkpoint: we do not duplicate them as individual variant results. The human comparison uses eighteen recordings and six interpreters; it does not establish universal superiority. Not reproduced by Inferama.
i Family results published; no independently verified scores per variant.
Inferama only highlights a “best result” when the metric, test set, configuration, and date allow for an equivalent comparison. The supplier's figures are presented as claims from its own source.
06 / VERSIONS
Chronology
Hy-SimulMT-30B-A3B-Interleave
Version added to Inferama's verified catalog.
07 / SOURCES