In light of the model degradation issues faced by users of the GLM Coding Plan, Zhipu has announced that it will offer compensation to those affected. In the same announcement, the company revealed that users will be granted additional opportunities for quota resets following the launch of the new model, indirectly signaling the impending release of the next-generation GLM-5.5. According to recent leaks, GLM-5.5 is rumored to have a parameter scale ranging from 2.5 trillion to 3.5 trillion, significantly surpassing the 743 billion parameters of its predecessor, GLM-5.3. As a Mixture of Experts (MoE) model, the actual number of parameters activated during inference will not increase proportionally with the total parameter growth. Furthermore, GLM-5.5 is expected to undergo substantial architectural enhancements, continuing to follow the technology path that balances capability and efficiency, as demonstrated in GLM-5.3-Flash. It may also incorporate the recurrent Transformer architecture, which Zhipu has preliminarily validated. This architecture can enhance the effective inference depth and improve the model's ability to handle multi-step complex tasks without a substantial increase in parameter count. Considering Zhipu's ongoing advancements in programming, agents, deep reasoning, and inference efficiency this year, there is high anticipation for the upgraded performance of GLM-5.5.