Perplexity has introduced a research version of a new orchestrator model for its local AI agent, Perplexity Computer, based on GLM 5.2 from China’s Z.ai and further trained for operation in the Computer environment. The model delivers near-frontier performance at 0.344x of the cost of Opus, according to Perplexity, and is intended to function in a hybrid scheme where GLM 5.2 handles most requests and escalates tasks to a more powerful model when necessary. Perplexity CEO Aravind Srinivas stated that when paired with an advisor, this model operates at Opus 4.8 level performance at significantly lower costs. The new system is already available as a research preview, and Perplexity has promised to release full benchmarks in the coming weeks. Perplexity Computer functions as an agent system capable of distributing tasks among multiple AI models, orchestrating over 19 models with the adapted GLM serving as a cost-effective base layer. GLM 5.2 contains about 744 billion parameters and is released under the MIT license. Z.ai has been on the U.S. Entity List since January 2025, and Perplexity hosts the adapted version in the U.S. on its own infrastructure.