Perplexity’s Deep Research uses DeepSeek R1, with limited free access

Perplexity has launched a Deep Research tool powered by a customised version of DeepSeek R1, offering people web-crawled reports rather than a quick answer and a list of links. Free users get a limited number of answers each day; the $20-a-month Pro plan includes unlimited access, according to Mashable.

Perplexity Watch analysis

What happened

The tool searches the web and compiles its findings into an in-depth research report. Mashable says Perplexity uses a customised version of the open-source DeepSeek R1, rather than a proprietary model of its own, as Google, OpenAI and xAI do for their respective research tools.

That model choice is the notable wrinkle: Perplexity is using an outside open-source model to power a flagship research feature. The service is available to free users with daily limits, while Pro subscribers get unlimited access as part of their subscription.

Why it matters

Research agents promise to turn a question into a report without making the reader do all the searching and stitching together. A free, limited tier also gives people a way to try one without first reaching for a paid subscription.

But a polished report is not the same thing as a dependable one. Mashable cites an example in which the tool wrongly attributed the phrase “stochastic parrots” to AI researcher Gary Marcus; it was coined by Emily M. Bender. The report also notes concerns about outdated or inaccurate information, especially in financial analysis.

Our read

Perplexity’s choice of DeepSeek R1 makes this more than another branded research button: it is a live test of how far a customised open model can carry a consumer research product. The free tier lowers the barrier to trying it, but treat the output as a starting point, not a signed-off report. Check the cited sources before relying on a claim, particularly where the stakes are financial.

What to watch

• Whether Perplexity improves source attribution and the freshness of its research.
• Whether the free daily limit is enough to make the feature useful beyond a trial.
• How the customised R1 version performs against competing research tools on accuracy, not just presentation. Activity teaser Perplexity’s new Deep Research tool uses a customised version of DeepSeek R1 to turn web searches into reports. Free users get limited daily access; Pro subscribers get unlimited use for $20 a month. That gives people a way to try an AI research agent without paying upfront, but Mashable notes an example of a misattributed phrase and concerns about stale or inaccurate information. A report with footnotes still needs a reader willing to check them. Is a free, limited research tool useful enough to trust with your first draft, or should these agents prove their accuracy before being treated as research assistants?

Discussion spark: Should AI research tools be judged mainly on how much searching they save, or should accurate attribution be a minimum requirement before their reports are useful?

not affiliated with or endorsed by Perplexity