Perplexity Shakes the AI World: Advanced Deep Research Claims #1 Spot
Just when we thought the AI race had reached a steady rhythm of incremental updates from the “Big Three” (OpenAI, Anthropic, and Google), Perplexity has thrown a massive wrench into the gears.
Today, February 7, 2026, Perplexity officially rolled out its Advanced Deep Research capabilities. The results aren’t just good; they’re industry-defining.
The Benchmark That Changed Everything
The headline story is the Google DeepMind Deep Search QA leaderboard. For months, we’ve seen a tug-of-war between OpenAI’s GPT-5 variants and Anthropic’s Opus 4.5. But as of today, a new king sits on the throne.
| Model / System | Score (%) |
|---|---|
| Perplexity Advanced Deep Research | 79.5% |
| Moonshot K2.5 | 77.1% |
| Anthropic Claude Opus 4.5 | 76.1% |
| OpenAI GPT-5.2 | 71.3% |
Perplexity isn’t just winning; it’s widening the gap. Scoring nearly 80% on a benchmark designed by DeepMind to test complex, multi-step research queries is a staggering achievement for a company that started as a search wrapper.
What Makes “Advanced” Deep Research Different?
This isn’t just a faster version of the search you’re used to. Perplexity has integrated several “next-gen” AI agent features:
- Evidence Graphs: Instead of just giving you a text summary, the tool now visualizes the research path. You can see exactly how it jumped from a financial report to a regulatory filing to a news article.
- Professional-Grade Citations: The citation engine has been rebuilt to meet the standards of legal and academic writing, providing direct links to specific pages within PDFs and datasets.
- Domain Optimization: The system now has specialized “modes” for Finance, Healthcare, and Legal research, utilizing specialized fine-tuned models to handle jargon and complex data structures.
- The DRACO Benchmark: In a move towards transparency, Perplexity has open-sourced DRACO, the benchmark they used to measure research accuracy, inviting the community to verify their claims.
Why This Matters for the “Big Three”
For OpenAI and Anthropic, this is a wake-up call. While they have been focused on raw model intelligence (like the recently released GPT-5.3 and Claude 4.6), Perplexity has focused on the workflow.
By combining high-end reasoning (using a mix of their own fine-tuned models and Anthropic’s Opus 4.5 engine) with real-time web access and synthesized reporting, Perplexity is replacing hours of human “manual labor” with a 2-to-4-minute automated process.
How to Access It
If you’re a Perplexity Pro or Max subscriber, the rollout has already begun. Pro users are reportedly getting a generous limit of 500 deep research queries per day, signaling Perplexity’s confidence in their infrastructure.
The AI race in 2026 is no longer just about who has the biggest LLM—it’s about who can build the most useful Agent. And right now, Perplexity is leading the pack.
Stay tuned to the Cbronai Blog for more updates on the rapidly evolving AI landscape.