Perplexity Unveils Hybrid AI System Using Zhipu's GLM 5.2, Signals Industry Shift to Cost Efficiency

According to Perplexity CEO Aravind Srinivas, this week the company unveiled a hybrid AI orchestration system that automatically selects the most suitable model for each task, rather than relying on a single large language model. The system leverages Zhipu's open-source GLM 5.2 model as its core architecture, routing routine tasks to lower-cost models and escalating complex operations to more powerful models only when needed.

The shift reflects a broader market transition from competing on model scale to competing on cost efficiency and task optimization. Over 85% of Fortune 500 companies have now adopted Ollama, a platform enabling enterprises to deploy and manage open-source AI models on-premises, according to Ollama CEO Jeff Morgan. Industry analysts predict that within 18 to 24 months, over 90% of AI tokens could be generated by open-source models rather than proprietary commercial systems.