FlashNEWS BRIEFING

iFlytek Launches Fully Domestic Speech Foundation Model Spark-Audio-1.0-Preview

1 min read

88Z AI, 17 September — iFlytek has released Spark-Audio-1.0-Preview, a speech foundation model trained on fully domestic computing power. Utilizing a 0.65B audio encoder and a 30B-A3B MoE language model trained on 13 million hours of audio, it directly understands speech semantics, emotions, and scenarios.

At a glance

  • Trained on fully domestic compute using 13 million hours of audio and MoE architecture
  • Supports 99 languages and 202 dialects, achieving SOTA on Fleurs Chinese dataset
Cover image for iFlytek Launches Fully Domestic Speech Foundation Model Spark-Audio-1.0-Preview
Cover image for iFlytek Launches Fully Domestic Speech Foundation Model Spark-Audio-1.0-Preview

The update

iFlytek has officially launched Spark-Audio-1.0-Preview, a fully domestic speech foundation model. Trained entirely on domestic computing power, the model combines a 0.65B-parameter audio encoder with a 30B-A3B Mixture of Experts (MoE) language model, leveraging 13 million hours of audio data for deep training. Functionally, the model directly understands semantics, emotions, and scenarios within speech. It supports recognition of 99 languages and 202 dialects, demonstrating leading performance in complex acoustic environments such as high noise and low volume. Furthermore, it achieved state-of-the-art (SOTA) results on the Fleurs Chinese dataset, showcasing robust speech processing capabilities.

Sources & editorial note

This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.

  1. 科大讯飞Source published

Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.

Related reading

All stories
88Z.ai