NativeAIHub

How Deep Research Works

All plans (limits vary)2 min read
1

Plan

Analyzes your question, identifies subtopics, and creates a multi step research strategy with specific search queries for each area.

2

Browse

Executes dozens of web searches across different angles, visiting multiple results per query across the open web.

3

Read

Reads full articles, documents, and PDFs. Interprets images, charts, and graphs. Extracts data and key findings from each source.

4

Reason

Evaluates findings for consistency, identifies contradictions, spots gaps in understanding, and generates new questions to investigate.

5

Pivot

Adjusts strategy based on what it has learned. Follows promising leads, abandons unproductive paths, and conducts targeted follow up searches.

6

Synthesize

Compiles all findings into a structured report with sections, summaries, key takeaways, and inline citations to every source.

0%

Humanity's Last Exam (vs o1: 9.1%)

0%

GAIA benchmark score

0 min

Maximum research session

Full Model (GPT-5.2)

Duration

10 to 30 minutes

Sources analyzed

Hundreds

Report depth

Comprehensive, multi section

Reasoning quality

Deep, multi step reasoning

Best for

Critical research, thorough investigations

Plus quota

10/month

Pro quota

125/month

Lightweight (o4-mini)

Duration

5 to 15 minutes

Sources analyzed

Dozens

Report depth

Concise overview

Reasoning quality

Faster, shallower reasoning

Best for

Quick overviews, less consequential queries

Plus quota

15/month

Pro quota

125/month

Use your full model queries wisely

On the Plus plan, you get only 10 full model (GPT-5.2) queries per month. Reserve these for research that truly requires depth: competitive analysis, academic reviews, due diligence, and complex technical investigations. Use lightweight (o4-mini) queries for less critical research where a quick overview is sufficient.