Perplexity Computer Adds Four Effort Presets to Control AI Task Depth
Perplexity Computer now lets users pick effort levels from Light to Ultra, and the orchestrator picks the right model and reasoning depth automatically.
- Perplexity Computer adds an effort slider with Light, Standard, High, and Ultra presets.
- Each preset bundles an orchestrator model choice with a reasoning depth setting.
- Lower settings use cheaper models; credit usage scales with actual work done.
- Custom mode still lets users pick a specific model and reasoning level manually.
- Live on web now, Android and iOS coming soon, per the announcement.
- Subagents can pull from multiple providers under a single orchestrator plan.
Perplexity Computer adds effort presets for agentic tasks
Perplexity has added four effort presets to Perplexity Computer, its agentic assistant. The control bundles model selection and reasoning depth into a single setting, allowing users to choose how much effort Computer should spend on a task.
The effort selector is available on the web, with Android and iOS support coming soon, according to Perplexity. It appears in the existing model selector inside the omnibar, where users can choose a preset before submitting a task.
Four presets, from routine to open-ended
| Preset | Best suited to | Example |
|---|---|---|
| Light | Routine, easy-to-check work | Collecting vendor invoices in a spreadsheet |
| Standard | Everyday research and multi-step workflows | Combining account notes and company updates into a sales brief |
| High | Complex analysis that informs decisions | Comparing customer requests with engineering estimates |
| Ultra | Open-ended analysis and long-running tasks | Evaluating a market under several growth assumptions |
One setting controls planning and depth
Each preset selects the orchestrator model that plans the assignment and sets how deeply it reasons before acting. The orchestrator breaks the task into steps and delegates work to supporting agents, which can use models from several providers.
Perplexity says it evaluates models at different reasoning levels and compares output quality with cost. The four presets package those benchmark results into defaults for common workloads. Lower settings use less expensive models, while final credit consumption also reflects the work Computer performs and delegates.
Users who need exact configuration can switch to custom mode and pin a specific model and reasoning level. That option preserves Computer’s previous workflow for repeatable tasks or cases where model choice is part of the requirement.
Orchestration moves to the foreground
The update continues Perplexity’s work on coordinating models across providers, including its earlier Model Council feature. Effort presets reduce the configuration required for ad hoc tasks, while custom mode gives developers and researchers direct control when they need a fixed model and reasoning depth.
Comments
