88Z AI, 21 October — Xingchen Lab has open-sourced the intent understanding model Xing4.0-29B-Intent, releasing 2.75 million training corpus entries, model weights, and inference images. The model uses a unified semantic parsing framework to support intent classification and parameter extraction, converting natural language into structured JSON. Evaluations show a comprehensive score of 58, surpassing the top score of 49 achieved by 18 mainstream general large models.
At a glance
- Open-sourced model with 2.75M training corpus, weights, and images
- Scored 58 in evaluations, beating the top score of mainstream general LLMs

The update
Xingchen Lab, under China Telecom Tianyi AI, has officially released the specialized intent understanding model Xing4.0-29B-Intent and announced its full open-source availability. This release includes not only model weights and inference images but also opens up a training corpus containing 2.75 million data entries, aiming to facilitate community development and optimization. Technically, Xing4.0-29B-Intent employs a unified pattern-driven semantic parsing framework capable of handling both intent classification and parameter extraction tasks simultaneously. It efficiently converts user natural language instructions into structured JSON format. According to official evaluation data, the model achieved a comprehensive score of 58, significantly outperforming the highest score of 49 among 18 mainstream general large models, demonstrating strong competitiveness in specific vertical domains.
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.
- 中国电信天翼AISource published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.