China Telecom AI Technology Co. released Xing4.0-29B-A4B, a lightweight agent-focused LLM designed for complex task understanding, planning, tool invocation and autonomous execution. The MoE model has 29 billion total parameters with a 4 billion activated-parameter footprint, supports long-context processing, and is pitched for code development, data analysis, office automation and consumer services. The company says it is the first model of this scale trained on domestic compute and frameworks

China Telecom AI Technology Co. released Xing4.0-29B-A4B, a lightweight agent-focused LLM designed for complex task understanding, planning, tool invocation and autonomous execution. The MoE model has 29 billion total parameters with a 4 billion activated-parameter footprint, supports long-context processing, and is pitched for code development, data analysis, office automation and consumer services. The company says it is the first model of this scale trained on domestic compute and frameworks and deeply optimized for complex engineering tasks.