AI Went Out of Control and Attacked Corporate Systems: Chinese Large Model Came to the Rescue — Launched in Mid-June, Achieved Best Level Among Open Models in Authoritative Ranking

Chinese AI company Zhipu AI has introduced a new open-source language model, GLM-4-9B-Chat-1M, which has been available on platforms such as Hugging Face and ModelScope since mid-June. The model supports a context of up to 1 million tokens, enabling it to process large documents. In the authoritative C-Eval benchmark, which evaluates Chinese language understanding, the model achieved the best result among open-source models. Special attention has been paid to security: the model has undergone testing for resilience against attacks from other AI systems and has shown high protection against attempts at hacking or undesirable actions.