Perplexity announced Decider for Mac on Oct. 22, an app it says runs its 4B pplx-decider-flash-v1.2 decision model on the device. Users can ask questions about text or images and get probabilities back, according to the company. In one documentation example, a question about a support message produces probabilities for routing it to billing, shipping or technical support.
What running it takes
The app serves a local Decisions API at http://127.0.0.1:8080/v1/decisions by default; users can change the port. Perplexity says local requests need no Perplexity API key and incur no usage bill. A request supplies content as state and a named question.
Perplexity lists Apple silicon (M1 or later), macOS 13 or later, 16 GB of memory and about 6 GB of free disk space as requirements. On first launch, the app downloads about 5.8 GB of model files. Downloading the app and model requires a network connection, and users must provide a Hugging Face token with read access if prompted.
Local processing and a hosted option
In a quote-post, @AravSrinivas claimed the model runs locally “without any data leaving your machine.” Perplexity's documentation says the model reads submitted content on the Mac, while noting the network connection needed for downloads.
Developers can move a request to Perplexity's hosted API by changing the URL and adding an Authorization header. Perplexity says the Mac's model weights are a q8_0 GGUF build, not the hosted deployment, and advises checking any locally tuned thresholds against hosted results before production use.
Comments