AI Industry Shifts From Model Size to Cost Efficiency, Perplexity Adopts Open-Source GLM 5.2

According to Perplexity CEO Aravind Srinivas in a CNBC interview, the AI industry is shifting from prioritizing model scale to cost efficiency and task-specific optimization. Companies are building orchestration systems that automatically select the most suitable model for each task rather than using the most powerful model for all applications. Customer service applications may not require expensive models, while complex programming tasks demand higher-tier models, and routine enterprise work can be handled by lower-cost open-source models.

Perplexity this week showcased a new computer-use system powered by Chinese AI company Zhipu's open-source model GLM 5.2, reflecting the trend toward hybrid AI architectures. Benchmark partner Peter Fenton predicts that over 90% of AI tokens could be generated by open-source models within 18 to 24 months. Ollama CEO Jeff Morgan revealed that over 85% of Fortune 500 companies now use Ollama's open-source model management platform, including highly regulated sectors like aviation, insurance, and healthcare.