Today, at the 2025 Zhongguancun Forum, Zhipu AI unveiled its latest Agent product – AutoGLM Contemplation. This is the industry's first Agent product that integrates deep research capabilities and operational capabilities, its key feature being its ability to "think and act simultaneously."
AutoGLM Contemplation breaks through the limitations of traditional AI models, enabling simultaneous complex thinking and execution. It can open and browse web pages like a human, completing the entire workflow from data retrieval and analysis to report generation, achieving a seamless integration of thought and action.
Zhipu stated that the technology behind AutoGLM Contemplation is Zhipu's fully self-developed GLM full-stack large model, mainly including: Inference model GLM-Z1-Air: Performance comparable to DeepSeek-R1, but with up to 8 times the speed and a cost of only 1/30th of DeepSeek-R1; Base model GLM-4-Air0414: Provides fundamental capabilities for the agent.
Zhipu further announced that the above models will be open-sourced on April 14th and will be gradually launched on the bigmodel.cn platform in the near future.
The launch of this product marks a significant step towards more practical and user-friendly agent technology, promising breakthroughs in research analysis, information processing, and other fields.
