Perplexity Deep Research

Perplexity · Research & Search

POST /v1/chat/completions

Research model for multi-step retrieval, synthesis, and reasoning, autonomously searching, reading, and evaluating sources across complex topics.

At a glance

Field	Value
Model id	`perplexity-deep-research`
Input modalities	Text, Image, Document
Output modalities	Text
Context window	128K
Weight precision	-
Features	web_search, reasoning, deep_research
Native inference	No
New	No
Supported endpoints	`POST /v1/chat/completions`, `POST /v1/responses`, `POST /v1/messages`, `POST /v1/search`

Pricing

Charge	Spec	Rate
Input	per 1M prompt tokens	$4.80
Output	per 1M generated tokens	$19.00
Citation Tokens	per 1M tokens	$4.80
Reasoning Tokens	per 1M tokens	$7.20
Search Queries	per query	$0.012

Example request

$ curl https://api.empiriolabs.ai/v1/search \
>   -H 'Authorization: Bearer $EMPIRIOLABS_API_KEY' \
>   -H 'Content-Type: application/json' \
>   -d '{"model": "perplexity-deep-research", "query": "latest LLM benchmarks 2026"}'

Parameters

Parameter	Type	Required	Default	Description
`temperature`	number	no	`0.2`	Sampling temperature. 0 = deterministic, 2 = maximum randomness. · Range: 0 – 2
`max_tokens`	number	no	`16384`	Maximum tokens in the response. · Range: 1 – 65536
`reasoning_effort`	enum	no	`"high"`	Level of reasoning depth. Higher effort = more thorough analysis but slower. · Allowed: `low`, `medium`, `high`
`search_mode`	enum	no	`"default"`	Filter search results by source type. Academic = scholarly papers, SEC = financial filings. · Allowed: `default`
`search_domain_filter`	string	no	-	Comma-separated domains. Prefix with ’-’ to exclude. Max 20.
`search_language_filter`	string	no	-	Comma-separated ISO 639-1 codes (2 letters). Max 10.
`search_after_date_filter`	string	no	-	Only include sources published after this date. Use YYYY-MM-DD or natural language.
`search_before_date_filter`	string	no	-	Only include sources published before this date. Use YYYY-MM-DD or natural language.
`search_recency_filter`	enum	no	`"none"`	Relative time filter. Cannot be combined with specific dates. · Allowed: `none`, `hour`, `day`, `week`, `month`, `year`
`last_updated_after_filter`	string	no	-	Only include sources last updated after this date.
`last_updated_before_filter`	string	no	-	Only include sources last updated before this date.
`country`	string	no	-	Two-letter ISO 3166-1 code (e.g. US, GB) for biasing results to a country.
`region`	string	no	-	State or province name for location bias.
`city`	string	no	-	City name for location bias.
`latitude`	number	no	-	Latitude for location bias, in decimal degrees.
`longitude`	number	no	-	Longitude for location bias, in decimal degrees.
`disable_formatting`	boolean	no	false	Skip the EmpirioLabs Markdown formatting (citation [N] rewriting + References block + <think> extraction to reasoning_content). Raw mode keeps the <think>…</think> tags inside content so callers can parse reasoning themselves.

Notes

Long-running multi-step retrieval and reasoning. Pricing splits Input / Output / Citation / Reasoning tokens plus per-query search charge.

Machine-readable schema: GET https://api.empiriolabs.ai/v1/models/perplexity-deep-research.