Perplexity
Perplexity’s Sonar Deep Research synthesizes broad web research into reports, with separate search, citation, and reasoning charges.
At a glance
API model ID: sonar-deep-research
Capabilities & boundaries
Tool support requires the appropriate API integration; a supported tool is not automatically active in every chat.
The Sonar model card documents reasoning_effort low, medium (default), and high. Higher effort increases research depth, reasoning tokens, latency, and cost. The current migration guide maps this model to the Agent API high preset; that replacement configuration is not the same control or a promise of an unchanged underlying model.
Put it to work
Original examples from EZ Ai Assist. Adapt these to your task and the features available in your workspace.
Workflow 01
Define scope before asking for an extensive report.
Prepare a research report on the software category below for a small product team. Limit the review to the stated geography and date range. Cover established products, emerging approaches, buyer requirements, and unresolved technical barriers. Cite primary sources, label vendor claims, and finish with an evidence table and research gaps.
Workflow 02
Make search coverage and exclusions inspectable.
Review publicly accessible research on this technical question. Explain your search scope and inclusion criteria, group the main approaches, and compare methods and reported limitations. Cite each study, distinguish peer-reviewed work from preprints, and identify contradictory findings. Do not invent papers or treat missing evidence as a negative result.
Workflow 03
Turn a broad review into concrete follow-up questions.
Research this technology stack for a potential adoption decision. Cover project maintenance, release history, licensing sources, operational requirements, and documented migration risks. Cite evidence with dates, separate facts from judgments, and end with questions for maintainers and a small proof-of-concept checklist. Do not make a final legal or security determination.
Developer reference
These are Perplexity Sonar API reference prices, not EZ Ai Assist subscription prices.
View EZ Ai Assist plans| Token type | Price |
|---|---|
| Input | $2.00 |
| Output | $8.00 |
| Citation tokens | $2.00 |
| Reasoning tokens | $3.00 |
| Usage | Rate and unit |
|---|---|
| Search queries | $5.00per 1,000 searches |
Common questions
Use deep-research-style workflows when a question warrants a broad source review and a structured report rather than an instant answer. Define the research boundary, date range, and evidence standard first, and budget for multiple searches and reasoning tokens.
Perplexity says Sonar Chat Completions support ended September 27, 2026. Synchronous and streaming calls continue through a gradual translation to Agent API requests; async Sonar calls are no longer supported. This does not establish whether EZ Ai Assist or a reseller exposes a particular model.
Perplexity maps Sonar Deep Research to high. Treat this as migration guidance, not an identical model or a guaranteed equivalent response. Recheck citations, quality, latency, and usage on representative tasks.
Searches, citation tokens, and reasoning tokens are separate metered items in addition to ordinary input and output. Research breadth affects work performed; inspect the response usage rather than guessing from prompt length.
No. Open the cited pages and confirm they support the exact claim, concern the correct version, and are recent enough. A citation is a starting point for verification, not a guarantee of accuracy.
The model card lists a 128K-token context. It does not establish a maximum output limit or training cutoff in the reviewed documentation. Integration limits may be lower, and search-backed answers still need checking.
No. The reference table describes Perplexity’s documented API billing. EZ Ai Assist plans are separate, and app availability must be checked in the workspace. These prompts are original starting points, not verified benchmark outputs.
Check the source
Specifications and API prices checked on . Example prompts and workflow advice are editorial guidance from EZ Ai Assist.
Same provider
Perplexity
Perplexity’s lightweight web-grounded Sonar model for focused questions, with citations and a documented transition to the Agent API.
Perplexity
Perplexity’s Sonar Pro supports deeper search and a 200K-token context. Explore its reference pricing and Agent API transition.
Perplexity
Perplexity’s Sonar Reasoning Pro combines multi-step analysis with web retrieval. Distinct from the retired sonar-reasoning API identifier.