Perplexity's Hybrid Compute splits sensitive tasks between cloud and local AI

Perplexity’s new Hybrid Compute lets users split a task between a cloud-based frontier model such as Opus 5 or GPT-5.6 Sol and a locally run LLM (Gemma E4B or Qwen 3.6 35B variants), keeping sensitive data on the Mac. The system automatically classifies confidential files, shows CPU/GPU usage, and charges only for cloud tokens, while requiring Apple Silicon Macs with macOS 15 and at least 32 GB unified memory.