NetEase Youdao fully open-sources Ziyue 4 multimodal model and speech synthesis model
Original title:网易有道“子曰 4”多模态模型、语音合成模型全量开源
AI summary
NetEase Youdao announced that the multimodal model and speech synthesis model of its Ziyue large model 4.0 are fully open-sourced globally. The multimodal model (27B parameters) focuses on education scenarios and reaches industry-leading levels in handling difficult visual math and science problems, with 81.4% accuracy on pure-text Chinese math and science problems. Through chain-of-thought optimization, the model compresses output length by 43.2%, effectively reducing inference costs. The simultaneously open-sourced speech synthesis model supports cross-lingual timbre and emotion transfer cloning, completing zero-shot replication within 3 seconds with over 97% accuracy, and supports 14 languages including Chinese, English, Japanese, and Korean.
Key Takeaways
- NetEase Youdao announced that the multimodal model and speech synthesis model of its Ziyue large…
- 0 are fully open-sourced globally.
- The multimodal model (27B parameters) focuses on education scenarios and reaches industry-leading…
- 4% accuracy on pure-text Chinese math and science problems.
- Through chain-of-thought optimization,the model compresses output length by 43.
Quick Facts
- Published
- Updated
- Source
- IT之家·人工智能
- Category
- Model Releases
- AI score
- 70
有道把垂直教育的多模态模型全量开源,27B参数在视觉数理上做到SOTA,还把思维链压缩了43%,推理成本实打实下降,做教育应用的可以拿来做二次开发;TTS的跨语种情感克隆也实用,3秒克隆14种语言。
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: NetEase Youdao fully open-sources Ziyue 4 multimodal model and speech synthesis model. May 22, 2026. https://new.nat6.net/en/news/cmpgrcly80hyfsljwa00hvczv/