Perplexity Computer adds local model support
Perplexity CEO Aravind Srinivas announced that Perplexity Computer will support running local models directly on personal hardware, including Macs and Linux machines. This update leverages local hardware for enhanced privacy and token efficiency per watt, while retaining the ability to route more demanding tasks to server-side GPUs running frontier models.
This hybrid compute strategy is the logical evolution for agentic platforms, optimizing for cost, latency, and privacy without sacrificing performance.
- –**Edge Efficiency**: Running smaller models on local hardware (like Apple Silicon) dramatically cuts down API costs and latency.
- –**Privacy Assurance**: Enterprise users will find local execution highly appealing for handling sensitive files and local data.
- –**Intelligent Routing**: The main engineering challenge will be building a seamless, low-overhead router that knows exactly when to call cloud-based frontier models.
DISCOVERED
1h ago
2026-06-02
PUBLISHED
1h ago
2026-06-02
RELEVANCE
AUTHOR
AravSrinivas