Perplexity announces hybrid compute splitting AI tasks, keeping data local
Perplexity announced Hybrid Compute, a new feature that lets you split AI tasks between powerful cloud models and your own device.
The big win here? Your sensitive data stays local, so privacy gets a major boost: some sensitive work can stay local.
Perplexity enables device or cloud processing
You can choose which files get processed and decide if they run on your device or in the cloud, using models like Gemma E4B or two versions of Qwen's 35-billion-parameter 3.6 model, one of which has been post-trained by Perplexity.
It's perfect for things like legal research where privacy matters most, plus it helps cut down on costs by letting your machine do some of the heavy lifting.
You also get a visualization of your system's handling things, CPU, GPU, memory, and all that techy stuff.
Installation is super easy (no terminal commands needed), and Perplexity plans to add even more local models.