Overview
Hy ASR 3.0 Preview is Tencent Hunyuan's August 2026 speech recognition update. It is positioned around context-aware transcription, meaning the model is designed to use more surrounding dialogue when converting speech to text.
This matters for long recordings where recognition quality depends on earlier terms, speaker context, or conversation structure. Buyers should evaluate it as a speech-to-text layer for meetings, calls, interviews, and media workflows rather than as a general AI agent or chatbot model.
What's New
Context-Aware Transcription
The central update is context-aware ASR behavior. Instead of handling each audio segment in isolation, Hy ASR 3.0 Preview is framed as a model that can better preserve meaning across longer speech spans. This can help with named entities, repeated terms, technical vocabulary, and punctuation decisions.
Better Fit For Long Conversations
The release is most relevant for meeting notes, customer service calls, internal interviews, podcasts, and livestream recordings. Those workflows often fail when ASR loses context between turns or misrecognizes terms that were introduced earlier.
Tencent Hunyuan Model Alignment
Hy ASR connects Tencent's speech recognition stack with the broader Hunyuan model family. That may matter for teams that want transcription, summarization, search, and voice-agent workflows to stay inside Tencent's cloud and product ecosystem.
Availability & Access
Hy ASR 3.0 is a preview release. Teams should confirm the official access path before production use, including whether the model is exposed through Tencent Cloud ASR, a Hunyuan API, Yuanbao, or another Tencent surface.
Preview status also means limits and pricing may change. For regulated audio, verify data retention, logging, region, and enterprise support before processing sensitive recordings.
Pricing & Plans
The reviewed release signal did not provide stable public pricing for the exact Hy ASR 3.0 Preview model. Tencent Cloud ASR products are typically evaluated through cloud usage terms, while consumer-facing access may be bundled differently.
| Plan or route | Pricing signal | Notes |
|---|---|---|
| Preview access | Confirm with Tencent | Availability may vary by product surface |
| Tencent Cloud ASR/API | Usage-based cloud billing likely applies | Check region, quota, and billing unit |
| Enterprise route | Contact Tencent | Best for private, high-volume, or compliance-sensitive audio |
Best For
- Teams transcribing long meetings, interviews, and customer conversations.
- Tencent Cloud users building voice analytics or speech-to-text pipelines.
- Product teams evaluating ASR for voice assistants or AI workflow inputs.
- Chinese-language organizations that need stronger contextual recognition.
- Developers who can test preview models before committing production traffic.
FAQ
What is new in Hy ASR 3.0 Preview?
The main update is context-aware speech recognition. The model is positioned to use broader dialogue context when transcribing longer audio.
Is Hy ASR 3.0 Preview generally available?
No. It is a preview release. Confirm the exact Tencent access route, support level, and pricing before production use.
Is this part of Tencent Hy3?
It belongs to Tencent's Hunyuan/HY model ecosystem, but it should be tracked separately from Hy3 because ASR is a speech recognition product line.
What should buyers test?
Use real recordings with overlapping speakers, background noise, technical vocabulary, timestamps, and downstream summary tasks.
Does Hy ASR replace audio cleanup tools?
No. Audio cleanup and ASR solve different problems. Cleanup can improve input quality, while ASR converts speech to text.
