Perplexity Attacks OpenAI, Launching Free Version of Deep Research

In partnership with

Shortly after OpenAI's launch of Deep Research, a tool that generates comprehensive research reports in minutes Perplexity have launched their own version, but free - undercutting OpenAI & Google who have been limiting the use of their version to paid plans.

Breaking Down the Paywall

Google and OpenAI have been increasingly putting their most advanced models behind paywalls. Google’s most advanced research features are locked behind a $20-per-month subscription, while OpenAI has imposed even steeper paywalls. OpenAI’s Deep Research feature is widely included in its $200/month ChatGPT Pro tier, with the $20/month ChatGPT Plus plan offering significantly lower usage gates.

Perplexity has opted for a radically different approach. By offering five free Deep Research queries daily to all users, it is positioning itself as a democratizing force in AI research.

The Compute Cost Factor

OpenAI’s decision to gate Deep Research behind a high-tier subscription is driven by the sheer computational expense of running these models. High-reasoning AI models like o1, o3, and OpenAI’s Deep Research require significantly more computing power per query than traditional models.

While the cost of raw compute is falling—dropping by roughly 90% every 18 months—the computational intensity of each query is increasing tenfold. OpenAI is using this dynamic to justify premium pricing and maximize lifetime value (LTV) from its consumer base.

Perplexity’s approach is different. While its research responses are generally shorter than OpenAI’s, it still delivers reasonably high-quality answers by scanning and aggregating hundreds of sources. The company leverages open-source models to keep its costs lower while maintaining competitive accuracy.

Performance and Benchmarking

Despite leveraging open-source models Perplexity’s Deep Research has shown impressive results in benchmarks:

• It scored 93.9% accuracy on the SimpleQA benchmark.

• It achieved 20.5% on Humanity’s Last Exam, surpassing Google’s Gemini Thinking and other leading models.

• OpenAI’s Deep Research remains the leader at 26.6% on Humanity’s Last Exam, but at a significantly higher cost.

Strategic Divergence: Perplexity vs. OpenAI

The core difference between Perplexity and OpenAI is structural. Perplexity does not develop its own foundational AI models but instead integrates and optimizes existing ones. This allows it to act as an aggregator, giving users access to multiple models tailored to different research needs. As a result, Perplexities aim is to commoditise the models it uses, with the aim of driving costs down to increase their own power and gross margin.

To further this goal Perplexity has recently introduced the ability to switch between models from multiple different companies, vs their previous UX that used a single model.

OpenAI, on the other hand, is focused on extracting the highest possible revenue from its user base. Despite its API offerings, OpenAI’s largest revenue stream comes from consumer subscriptions.

The number of ChatGPT subscriptions has grown rapidly, with the company tripling its ChatGPT Plus subscribers to 15 million. Its real growth focus, however, is on the $200/month ‘Pro’ tier.

A $200/month price point is a tough sell for many consumers, however, especially outside the U.S., unless they can directly justify the cost through increased productivity in a work context. 

The Next Battleground: Data Access

While reasoning models like Deep Research are becoming more powerful, they still struggle to generate truly novel insights. This is because they primarily rely on publicly available, generic information.

Today, to make the output usable, white collar professionals still need to provide models with the truly valuable and unique info—such as industry reports or proprietary financial analyses—remains locked behind expensive content paywalls.

Whilst this is fine for enterprises with large multi-seat subscriptions to dozens of content providers it still requires the ‘analyst’ to collate all the information and feed it to the model taking hours, if not days and doesn’t address the wider market that doesn’t have these subscriptions.

Therefore, to make these reasoning models truly ‘work’ the opportunity is in striking content deals that give the model access to high quality info from the likes of Bloomberg, Statista, CB Insights etc.

OpenAI have looked at a tangential approach, striking content licensing deals with publishers like the New York Times but this was (1) to avoid lawsuits, and (2) improve the quality of the written output of models - not providing access to real-time, gated expert info.

This places Perplexity, OpenAI or whichever company that does this as an aggregator, giving paying subscribers fractional access to massively high value content - instead of having to pay for dozens of subscriptions, improving the output and strongly justifies a $200/month price point (would cost much more to buy yourself).

This positions the company like Netflix in 2015 where you’d pay one $10/month subscription to get access to all Disney, Hulu, HBO Max, Peacock, and Paramount shows. 

This represents differentiated angle of attack for the larger, closed-source AI players giving their models access to real-time data open source models wouldn’t have.

Additionally, this strategy would counteract the playbook of model distillation—a process that compresses expensive AI models into more efficient versions—that currently threatens to commoditize frontier models as the data they’re training on would be immediately out of date.

Stay up-to-date with AI

The Rundown is the most trusted AI newsletter in the world, with 1,000,000+ readers and exclusive interviews with AI leaders like Mark Zuckerberg, Demis Hassibis, Mustafa Suleyman, and more.

Their expert research team spends all day learning what’s new in AI and talking with industry experts, then distills the most important developments into one free email every morning.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses – tailored to your needs.

Sign up to start learning.