Perplexity Computer Adds Effort Mode for Model Selection
Perplexity added effort controls to Computer on September 17, 2026, giving users a simple slider to set how much computational effort a task deserves instead of picking a specific underlying model.
Details
- Four effort levels: Light, Standard, High, and Ultra, each pairing a preset orchestrator model with a matching reasoning depth
- Light: aimed at straightforward, everyday tasks such as collecting vendor invoices
- Standard: balances reasoning and cost for tasks like combining account notes into a brief
- High: targets complex analysis, such as weighing customer requests against engineering estimates
- Ultra: applies maximum effort to open-ended, harder problems
- Automatic model matching: Perplexity selects the model and reasoning level for each effort tier itself, saying its experience answering billions of queries informs how it matches models to work
- Manual override still available: users who prefer to pick a specific model and reasoning level directly can still do so through custom controls
- Rollout: available now on web, with Android and iOS support coming soon
- Billing: credit usage scales with the effort level selected and the work actually performed
What happened next
The change responds to user feedback that people wanted control over how much a task should cost and how much effort it deserved, without needing to understand which underlying model to pick for each job. It follows Perplexity’s September 14 Portable Computer launch for Windows, continuing a run of usability refinements to the Computer product rather than a new model or platform release. Mobile support for effort mode had not shipped as of this writing.